Introduction To Data Management and Programming in SAS

BIO113 i
Introduction to Data Management and Programming in SAS
Harvard School of Public Health
Text book
The Little SAS Book (Second Edition), Delwiche LD and Slaughter SJ, SAS Institute, Inc.,
1998. Available at the Harvard Coop. Identified as LSB.
Reference
SAS User's Guide: Basics and SAS User's Guide: Statistics.
SAS System Help (online)
General Information (LSB 1.1-1.10, 3.11, 8.1-8.2)
What is as SAS data set ...............................................................................................1

What is as SAS program...............................................................................................1
SAS names ...................................................................................................................3
SAS variables ...............................................................................................................3
Lists of variables ...........................................................................................................3
SAS statements ............................................................................................................3
Coding style ..................................................................................................................4
SAS Windowing Environment .......................................................................................4
Creating SAS data sets (LSB 2.1-2.7, 2.12-2.16, 8.3-8.6, 8.13-8.14)
Begin the data step: DATA.......................................................................................... 10

Where's the data: DATALINES, INFILE ...................................................................... 10
Describe the 'raw' data to SAS: INPUT
List, column, formatted and mixed input statements ........................................ 11
INFORMAT (special formats) ........................................................................... 11
Delimited files ................................................................................................... 12
Pointer commands: @, +n................................................................................ 15
Conditional input ......................................................................................................... 18
Multiple records/observation, multiple observations/record......................................... 20
Options on INFILE: MISSOVER, TRUNCOVER, LRECL............................................ 21
BIO113 ii
Accessing external raw data sets and SAS data sets (LSB 2.8-2.10)
External raw data sets: FILENAME and fileref ............................................................ 26

SAS libraries ............................................................................................................... 27
Temporary vs. permanent SAS data sets ................................................................... 28
Create a temporary SAS data set ............................................................................... 28
Create a permanent SAS data set: LIBNAME and libref............................................. 30
Using a permanently stored SAS data set .................................................................. 31
Another way to create and later use a permanent SAS data set................................. 32
See what's in the SAS data set: PROC CONTENTS, Explorer/Libraries .................... 34
Input from other programs: Stat/Transfer™ ................................................................ 36
Create new variables and change existing variables (LSB 3.1-3.3, 3.7-3.8, 8.7-8.9)
Assignment statements............................................................................................... 37
Create new variables .................................................................................................. 37
Change existing variables ........................................................................................... 38
SAS functions .............................................................................................................39
Date functions ............................................................................................................. 40
Year 2K ....................................................................................................................... 41
Date difficulties............................................................................................................ 42
Useful SAS functions .................................................................................................. 43
Conditional processing (LSB 3.4-3.6)
Conditional assignment: IF-THEN, IF-THEN-DO-END ............................................... 44

Multiple choice processing: IF-THEN-ELSE................................................................ 46
Subsetting IF............................................................................................................... 47
Arrays (LSB 3.10)
Define arrays: ARRAY ................................................................................................ 48

Process arrays: DO...END .......................................................................................... 48
Create observations.................................................................................................... 53
Sorting and printing data with PROCs (LSB 4.1-4.7)
Basics: PROC PRINT, PROC SORT .......................................................................... 55

Dress up the output: LABEL, TITLE, FOOTNOTE, FORMAT ..................................... 57
Create you own permanent format library................................................................... 59
BIO113 iii
Simple data summarization (LSB 4.9-4.11, 4.14, 7.2-7.3)
Summarize categorical data: PROC FREQ ................................................................ 61

Summarize continuous data: PROC MEANS.............................................................. 64
Display relationships between variables: PROC PLOT, PROC GPLOT ..................... 68
Data management I - SET (LSB 5.1-5.3)
Basics .........................................................................................................................71
Concatenate 2 SAS data sets ..................................................................................... 72
Interleave 2 SAS data sets: BY................................................................................... 72
Create a subset: IF ..................................................................................................... 73
Data management II - MERGE (LSB 5.4-5.7, 5.9-5.10, 5.14)
Basics .........................................................................................................................74
One-to-one join ........................................................................................................... 75
One-to-many join ........................................................................................................ 76
Join with IN= option..................................................................................................... 77
"Many-to-one" join (doesn't use MERGE) ................................................................... 79
Data management III - OUTPUT (LSB 5.11) ....................................................................... 80
Data management IV - Change unit of observation (LSB 3.9-3.10, 5.12-5.14)
Make several observations from one .......................................................................... 81

Make one observation from several ............................................................................ 82
Data analysis - PROC step (LSB 7.1-7.8, SAS User's Guide: Statistics)
PROC UNIVARIATE ................................................................................................... 83

PROC TTEST .............................................................................................................85
PROC NPAR1WAY .................................................................................................... 86
PROC REG................................................................................................................. 87
PROC ANOVA ............................................................................................................ 89
PROC GLM................................................................................................................. 91
PROC LOGIST ........................................................................................................... 92
PROC LIFETEST........................................................................................................ 94
BIO113 1
GENERAL INFORMATION (LSB 1.1-1.10, 3.11, 8.1-8.2)
What is a SAS data set? a rectangular array of observations and variables with documentation.
Variables (columns)
Documentation
Observations (rows)
What is a SAS program? a series of SAS statements, each terminated by a semicolon (;),
executed in order on each observation (an implied loop with implicit output).
Components: OPTIONS statements

Data set definition (LIBNAME and FILENAME) statements
DATA step(s)
PROC step(s)
Comments
Run statements
OPTIONS: begins with keyword OPTIONS

governs the appearance of output
Data set definition: begins with the keyword LIBNAME or FILENAME

tells SAS where to find the data and where to write data
DATA step: begins with the keyword DATA

accesses 'raw' data or existing SAS data set(s)
modifies accessed data
creates a new SAS data set(s)
PROC step: begins with the keyword PROC

takes an action or performs an analysis
frequently generates a report
Comments: annotate and document SAS programs
Run: execute preceding lines in DATA or PROC step

BIO113 2
Example of a SAS program:

* example of a SAS program;
options nocenter;
data sasclass;
input name $ height;
datalines;
roland 180
rhonda 172
;
run;
proc print;
run;
SAS LOG
1 * example of a SAS program;

2 options nocenter;
3 data sasclass;
4 input name $ height;
5 datalines;
NOTE: The data set WORK.SASCLASS has 2 observations and 2 variables.

NOTE: DATA statement used:
real time 0.35 seconds
cpu time 0.13 seconds
8 ;
9 run;
10
11 proc print;
12 run;
NOTE: There were 2 observations read from the dataset WORK.SASCLASS.

NOTE: PROCEDURE PRINT used:
SAS OUTPUT
The SAS System
Obs name height
1 roland 180
2 rhonda 172
BIO113 3
SAS names: You, the SAS programmer, get to name variables, arrays and data sets but there
are a few rules.
Names consist of 32 or fewer letters, numbers or underscores (_) EXCEPT dataset

(member) names are ≤ 28 characters and librefs, filerefs, and formats are
≤ 8 characters
Names must begin with a letter or an underscore
Names may be lower or upper case (SAS doesn't recognize that upper and lower case are
different)
Examples:
SAS variables: variables are either character or numeric.
TYPE WHAT LENGTH MISSING VALUE

character everything 1-32767 blank
numeric 0, 1, 2...9 1-32 . (dot)
-,+, decimal point
scientific notation (E)
Lists of variables: (LSB 3.11)
regular list: name age sex wt bp1 bp2 bp3 bp4 bp5
'type' list: _numeric_, _character_
single dash list: bp1-bp5
double dash list: name--bp5
SAS statements: the building blocks of the SAS program. A few rules here, too.
Statements (except assignment statements) begin with a SAS keyword

Every statement ends with a semicolon (;)
Words in statements are separated by one or more spaces or by operators such
as +, -, *, /, <, <=, >, >=, =, ^=, &, |, lt, le, gt, ge, eq, ne, and, or
Statements may be lower or upper case
Statements may continue over multiple lines
Statements may start in any column
Most statements are unique to the DATA step (SET, MERGE, IF) or to the PROC step
(VAR, TABLES, MODEL). A few appear in both (BY, LABEL, FORMAT, TITLE,
RUN, OPTIONS)
BIO113 4
Coding style: Develop a style that works for you. The idea is have a program that you and
people you work with can read and debug easily. Some suggestions...
Begin each SAS statement on a new line (but, I don’t always...)
When the statement extends beyond a single line, indent succeeding lines
Separate tasks in the program with blank lines
Use comments: * comment; or /* comment */
Complete each DATA and PROC step with a RUN statement
SAS Windowing Environment: Consists of five windows.
PROGRAM EDITOR: enter and modify SAS programs and data

execute programs
LOG: SAS's annotation of programs after execution (black)

has NOTES about data set names, number of observations and
variables, execution time (blue)
displays WARNINGS (green)
displays ERRORs (red)
OUTPUT: Results of PROCedures
RESULTS: Table of contents/index of your output
EXPLORER: Access to data libraries and files

BIO113 5
1. Start SAS
Click Start, Statistics and SAS8 to open the SAS Windowing Environment. Three windows
(Editor, Log, Explorer) are visible when you start. Output and Results may be accessed by
clicking the appropriate tab at the bottom of the screen.
The title bar of the active window is blue while the others are gray. Click anywhere in a
window to make it the active window.
BIO113 6
2. The SAS menu bar
File open, save and print files and exit SAS

Edit editing commands such as cut, copy and paste, and search/replace
View select the active window
Tools various editors and SAS options/preferences and options/'keys' settings
Run submit and recall SAS programs
Solutions some point-and-click programs (may not be available)
Window select tiled or cascaded windows, select active window
Help access SAS on-line help
Notice the Run item on the menu bar disappears when the editor window is NOT the active
window.
3. The SAS tool bar (command bar and various buttons)
Command submit short commands to SAS

buttons shortcuts to actions you may also access from the menu bar. Move the cursor
over a button and a short description appears. The button at the right end, for
example, accesses SAS help.
4. The SAS window bar
Click on a tab to see a hidden window and make it active.
5. The SAS status bar
The status bar is at the bottom of the screen. Error messages will appear in the area
on the left. The area on the right displays the current working folder (directory). You may
change the working folder by double-clicking this area and navigating through folders in the
usual Windows way (we will demonstrate this in class).
BIO113 7
6. Enter and save a SAS program

Make the p:\bio113 your current working folder and make the editor window active (click
anywhere in the window). Click on the 1 in the upper right corner of the window to make it
fill the screen. Type in the following program. The editor is a primitive word processor—
arrow keys move you around, the End key moves you to the end of a line, Home moves you
to the beginning of a line, PageUp and PageDown move you up or down one page, Enter
moves you to the beginning of a new line. You can block areas of the program and cut and
paste in the usual way. Pay close attention to spelling and punctuation as you type the
program. If you misspell a SAS key word, it will be colored red until you correct it. Notice
that comments are green, key words are blue, data in the program are highlighted in yellow,
title text is purple, and everything else is black. In addition, rules appear to separate parts of
the program (data and proc steps). It's a very good idea to save your program before you
submit. It can be recalled from memory during this SAS session but, should you exit SAS
before saving it, it will be lost forever. Click File and Save and give the file the name
first.sas. Because your working folder is p:\bio113, that is where the file will be saved.
By the way, you could click on the save button and fill in the dialogue to save the file.
BIO113 8
6. Submit the SAS program for processing

Click Window and Tile (vertical or horizontal) so that the editor, log and output windows are
visible and then on the little man in the tool bar. Text will whiz by in the LOG window
and, if the program runs successfully, output will appear in the OUTPUT window.
7. Examine the LOG window first

Make the LOG window active and click 1 to make the window large.
Your SAS program and messages from SAS are displayed in different colors (you'll see the colors
in class).
Black the program you submitted
Blue SAS notes--number of observations and processing time
Green warnings
Red errors
Green and red messages suggest that you should examine your program carefully for program
and/or logical errors. We'll review recalling, correcting and resubmitting problem programs in
class.
BIO113 9
8. Examine the OUTPUT window

Make the output window active (click Output in the window bar) and click 1 to make it large.
Scroll through the output with PageUp and PageDown or by moving the bar on the right side
of the window.
9. Print the contents of any window
Make the window you want to print active. Click File, Print and OK or click (the printer
button) on the tool bar.
10. Save the contents of the LOG and OUTPUT windows
Make the window you want to save active, click File and Save as or click on (the save
button) and fill in the dialogue. Name the LOG file first.log and the OUTPUT first.lst. Files
saved in this way may be imported into your favorite word processing program for formatting
and editing (notice that you cannot edit in the LOG or OUTPUT windows). You may also use
the usual copy and paste buttons or key strokes to copy LOG or OUTPUT material to a
word processing document.
BIO113 10
CREATING SAS DATA SETS (LSB 2.1-2.7, 2.12-2.16, 8.3-8.6, 8.13-8.14)
Begin the DATA step: DATA
Start with the keyword, DATA

Follow with your chosen data set name (if you don't, SAS will assign the name data1 to the
first data set, data2 to the second, etc.)
End the statement with ;
Example: data;
data whatever;
Where's the data? DATALINES and INFILE statements tell SAS where to find 'raw' data
DATALINES: The data are in the program
INFILE: The data are in an external file
Example:
* dataliness example; * infile example; external file (ex01.dat)

data one; data one; 11 35
input a b; infile 'ex01.dat'; 95 7
datalines; *infile 'p:\bio113\ex01.dat';
11 35 *infile 'g:\shared\bio113\ex1.dat';
95 7 input a b;
; run;
run; proc print;
proc print; run;
run;
Result (both programs produce this output):
The SAS System 15:15 Sunday, August 15, 2004 1
Obs a b
1 11 35
2 95 7
BIO113 11
Describe the 'raw' data to SAS: The input statement begins with the SAS keyword INPUT
and ends with a semicolon (;).
LIST input may be used if:
1. Data values are separated by one or more spaces or by a delimiter such as a tab or
comma
2. Character data contain no imbedded spaces and are 8 or fewer characters
3. Missing data values are indicated by a dot (.)
4. There's nothing funny like a date (if so, use an INFORMAT statement, too)
5. You plan to read every variable on the data line
How? List the names you have chosen for the variables
Indicate character variables by following their names with $
Example program:
(♦ in data lines indicate blanks)
data one; Data:

infile 'p:\bio113\ex02.dat';
input name $ age; sam♦42
run;
proc print;
run;
Result:
Obs name age
1 sam 42
data one; Data:

input id a b c; 21♦1♦123♦4456
run;
proc print;
run;
Result:
Obs id a b c
1 21 1 123 4456
BIO113 12
data one; Data:

input mrn bp1-bp5; 90043♦58♦67♦98♦72♦71
run;
proc print;
run; Result:
Obs mrn bp1 bp2 bp3 bp4 bp5
1 90043 58 67 98 72 71
data one; Data:

informat dob mmddyy10.; sam♦01/16/1960
input name $ dob; william♦01/18/1960
run;
proc print;
run; Result:
Obs dob name
1 15 sam
2 17 william
data one; Result:

informat name $8. Obs name dob
dob mmddyy10.;
input name dob; 1 sam 15
run; 2 william 17
proc print;
run;
data one; Result:

input name $ dob : mmddyy10.; Obs name dob
run;
proc print; 1 sam 16JAN1960
format dob date9.; 2 william 18JAN1960
run;
data one; Data:

infile 'p:\bio113\ex06.dat' dlm=',';
input x y z; 2,5,41
run; 11,2,54
proc print;
run; Result:
Obs x y z
1 2 5 41
2 11 2 54
BIO113 13
COLUMN input may be used if:
1. Data values for each variable are always in the same columns in all the data lines in
the raw data set
2. There's nothing funny like a date (INFORMAT may not be used with column input)
Follow each name with the column location of the variable
Indicate character variables with $ before the column location
Example program:
data one; Data:

input name $1-10 age 11-12; sam♦♦♦♦♦♦♦42
run; alexander ♦12
proc print;
run;
Result:
Obs name age
1 sam 42
2 alexander 12
data one; Data:

input id 1-2 a 3 b 4-6
311123♦4456
c 8-11;
run;
proc print;
run; Result:
Obs id a b c
1 31 1 123 4456
BIO113 14
data one; Data:

input mrn 1-5 bp1 9-10
90043♦♦♦5867987271
bp2 11-12 bp3 13-14
bp4 15-16 bp5 17-18;
run;
proc print;
run;
Result:
1 90043 58 67 98 72 71
data one; Data:

input name $1-3 bm 4-5 pam01♦♦19♦♦1960
bd 8-9 by 12-15;
dob=mdy(bm,bd,by);
run;
proc print;
run;
Result:
Obs name bm bd by dob
1 pam 1 19 1960 18
BIO113 15
FORMATTED input may be used if:
1. Data values for each variable are always in the same columns in all the data lines in the
raw data set
2. There are funny variables like dates
3. You want to reuse formatting instructions
Follow each name with the format instruction for the variable (note the
'.' at the end of the format instruction)
Indicate character variables with $ before the format instruction
Use pointer instructions (@n and +n) to move to correct columns
Example program:
data one; Data:

input name $10. age 2. sam♦♦♦♦♦♦♦4201161960
dob mmddyy8.;
run;
proc print;
run; Result:
Obs name age dob
1 sam 42 15
Data:
data one;
infile 'p:\bio113\ex12.dat'; 311123♦4456
input id 2. a 1. b 3. @8 c 4.;
run;
proc print;
run; Result:
Obs id a b c
1 31 1 123 4456
BIO113 16
data one; Data:

input mrn 5. +3 (bp1-bp5)(2.);
run; 90043♦♦♦5867987271
proc print;
run;
Result:
1 90043 58 67 98 72 71
data one; Data:

input @1 (x1-x3)($1. +3) a342d165f209
@1 (y1-y3)(+1 1. +2) m111x893p585
@1 (z1-z3)(+2 2.);
run;
proc print;
run;
Result:
Obs x1 x2 x3 y1 y2 y3 z1 z2 z3
1 a d f 3 1 2 42 65 9
2 m x p 1 8 5 11 93 85
Compare the above to this:
data one;
input x1 $1. y1 1. z1 2. x2 $1. y2 1. z2 2. x3 $1. y3 1. z3 2.;
run;
proc print;
run;
Result:
Obs x1 y1 z1 x2 y2 z2 x3 y3 z3
1 a 3 42 d 1 65 f 2 9
2 m 1 11 x 8 93 p 5 85
BIO113 17
MIXED input:
In most situations, you'll mix input styles, using what's convenient and requires the
least typing. Typically, you use list style or you use a mix of column and formatted
style. This example is a bit contrived to show that you might use all three styles.
Example program:
data one;
input name $ age mrn 14-18 dob mmddyy8.
+1 sex $1. @33 (bp1-bp5)(2.);
run;
proc print;
run;
Data:
sam♦♦♦♦♦43♦♦♦9004301161960♦m♦♦♦♦5867987271
pam♦♦♦♦43♦♦♦♦9011301191960♦f♦♦♦♦6165807870
Result:
Obs name age mrn dob sex bp1 bp2 bp3 bp4 bp5
1 sam 43 90043 15 m 58 67 98 72 71
2 pam 43 90113 18 f 61 65 80 78 70
BIO113 18
Conditional input may be used if:
1. You want to read some, but not all, data lines.

2. You want to read different data lines with different input statements.
How? Input the conditional variable and end the statement with @ (trailing at)
Test the value of the conditional variable
Execute additional input statements or delete the observation based on the
value found
Example program: Read some, but not all, data lines.
data one; Data:

input grade @; 6 bob m 11
8 bill m 13
if grade < 8 5 jane f 10
then input name $ sex $ age;
else delete;
run; Result:
proc print; Obs grade name sex age

run;
1 6 bob m 11
2 5 jane f 10
* read parent data; Data:

data one;
infile 'p:\bio113\exfam.dat'; house 8 5 3 2
input rec $ @;
parent 1 33 1
if rec='parent'
then input sex age employed; parent 0 33 0
else delete; child 1 10
child 0 9
run; child 1 6
pet 0 1
proc print; pet 0 2
run;
Result:
Obs rec sex age employed
1 parent 1 33 1
2 parent 0 33 0
BIO113 19
Example program: Read different data lines with different input statements.
data one;
input sex $6 @;
if sex='f'
then input name $1-4 @7 (softball fldhcky)(1.) age 9-10;
else if sex='m'
then input name $1-4 @7 (baseball soccer)(1.) age 10-11;
else delete;
Data:
run; 123456789012
proc print;
run; bob m11 10
jane f01 9
jill f1012
bill m01 12
lee 11
Result:
Obs sex name softball fldhcky age baseball soccer
1 m bob . . 10 1 1
2 f jane 0 1 9 . .
3 f jill 1 0 12 . .
4 m bill . . 12 0 1
* read all data; Data:

data one;
infile 'p:\bio113\ex16fam.dat'; house 8 5 3 2
input rec $ @; parent 1 33 1
if rec='house' then input rooms num kids pets; parent 0 33 0
else if rec='parent' then input sex age employed; child 1 10
else if rec='child' then input sex age; child 0 9
else if rec='pet' then input sex kind;
else delete;
child 1 6
run; pet 0 1
pet 0 2
proc print;
run;
Result:
Obs rec rooms num kids pets sex age employed kind
1 house 8 5 3 2 . . . .
2 parent . . . . 1 33 1 .
3 parent . . . . 0 33 0 .
4 child . . . . 1 10 . .
5 child . . . . 0 9 . .
6 child . . . . 1 6 . .
7 pet . . . . 0 . . 1
8 pet . . . . 0 . . 2
BIO113 20
Read multiple observations on a single record -- double trailing at (@@):
data one;
infile 'p:\bio113\ex18.dat'; Data:
input id hr @@;
run; 1001 121 1005 93 1011
proc print; 107 1034 82
run;
From SAS LOG:
NOTE: 2 records were read from the infile 'p:\bio113\ex18.dat'.

The minimum record length was 11.
The maximum record length was 21.
NOTE: SAS went to a new line when INPUT statement reached past
the end of a line.
OUTPUT:
Obs id hr
1 1001 121
2 1005 93
3 1011 107
4 1034 82
Read a single observation from multiple records – slash (/) or pound sign (#) pointer
instruction :
data one; Data:
input id 4. hr1 2. / hr2 2-3 #3 hr3; 102195
run; f88
proc print; 101
run;
From SAS LOG:

NOTE: The data set WORK.ONE has 1 observations and 4 variables.
OUTPUT:
Obs id hr1 hr2 hr3
1 1021 95 88 101
BIO113 21
Options on the INFILE statement:
With no options on the infile statement, we get into trouble trying to read the following data
with a list input statement. Notice, there are only 3 data values on the first data record.
Example program:
* infile with no options;
data one;
Data:
input bp1-bp4;
run; 121 83 100
proc print; 84 92 120 95
run;
Result:
From SAS LOG:

the end of a line.
OUTPUT:
Obs bp1 bp2 bp3 bp4
1 121 83 100 84
BIO113 22
MISSOVER If the list input statement describes variables beyond the end of the data
record, as above, SAS will go to the next line to complete the input statement.
MISSOVER tells SAS to stop reading variables at the end of the line; any variables in
the input statement not read before SAS finds the end of the line are assigned
missing (.). (Note: TRUNCOVER, page 23, will achieve the same effect but the
tradition is to use missover with list input and truncover with column and formatted
input)
Example program: Data:

* infile with missover - data from ex20; 121 83 100
data one; 84 92 120 95
infile 'p:\bio113\ex20.dat' missover;
input bp1-bp4;
run;
proc print;
run;
Result:
From SAS LOG:

OUTPUT:
Obs bp1 bp2 bp3 bp4
1 121 83 100 .
2 84 92 120 95
BIO113 23
With no options on the infile statement, we get into trouble trying to read the following data
with a column or formatted input statement. You cannot ordinarily see that a record is
"short", but here I've marked the end of the data record with .
Example program: Data

* infile with no options;
data one; 45385
763943
input x 1-3 y 4-6;
run;
proc print;
run;
Result:
From SAS LOG:

the end of a line.
OUTPUT:
Obs x y
1 453 763
BIO113 24
TRUNCOVER If the column or formatted input statement describes variables beyond the
end of the data record, as above, SAS will go to the next line to complete
the input statement. TRUNCOVER tells SAS to read to the last column
specified in the input statement or to the end of the line, whichever
happens first. By the way, MISSOVER will not work here.
* infile with truncover; 45385

data one; 763943
infile 'p:\bio113\ex21.dat' truncover;
input x 1-3 y 4-6;
run;
proc print;
run;
From SAS LOG:

OUTPUT:
Obs x y
1 453 85
2 763 943
LRECL Short for Logical RECord Length. Somewhere deep inside SAS it is written that
data lines (records) are of maximum length 256 unless otherwise specified. If the
data lines are longer than 256 characters, then you must describe the longer
record length with LRECL in order to read the data beyond column 256.
Example: infile 'loooong.data' lrecl=2001;

BIO113 25
PAD PAD, used with column or formatted input, will 'pad' the data line with spaces out
to the LRECL. This is another way to even out the 'ragged' right side of a data set
(you would not need to use truncover).
NOTE: You may end up using PAD on your infile statement but it is NOT a good idea to
use it unless it's needed. Ragged data lines in formatted files often indicate some
problem in the creation or copying of the data set. The 'SAS went to a new line'
message will warn you to look at the data file more carefully.
Example: infile 'ex21.dat' pad;

infile 'loooong.data' lrecl=2001 pad;
* infile with pad; 45385

data one; 763943
infile 'p:\bio113\ex21.dat' pad;
input x 1-3 y 4-6;
run;
proc print;
run;
From SAS LOG:

NOTE: The data set WORK.ONE has 2 observations and 2 variables
OUTPUT:
Obs x y
1 453 85
2 763 943
BIO113 26
ACCESSING EXTERNAL RAW DATA SETS AND SAS DATA SETS (LSB 2.8-2.10)
External raw data sets: FILENAME and fileref
Data:
We've seen:
11 35
* what we've seen; 95 7
data one;
infile 'p:\bio113\ex01.dat'; * note: name of external raw data set in single quotes *;
input a b;
run;
You may also describe the external 'raw' data set with a FILENAME statement. A link is
created between the file named in the FILENAME statement and the INFILE statement with
the fileref. You choose the fileref but it can only be 8 characters long. The FILENAME
statement must precede the INFILE statement. Options (MISSOVER, TRUNCOVER,
LRECL, PAD) appear on the INFILE statement as before.
How? FILENAME fileref 'raw-data set-name';
Example:
* filename with fileref (fluffy is the fileref);

filename fluffy 'p:\bio113\ex01.dat';
data one;
infile fluffy;
input a b;
run;
SAS LOG:
267 * filename with fileref (fluffy is the fileref);

268 filename fluffy 'p:\bio113\ex01.dat';
269 data one;
270 infile fluffy;
271 input a b;
272 run;
NOTE: The infile FLUFFY is:

File Name=p:\bio113\ex01.dat,
RECFM=V,LRECL=256
NOTE: 2 records were read from the infile FLUFFY.

BIO113 27
Test: Write FILENAME and INFILE statements to read the 'raw' data set, ivh.dat, from
p:\bio113.
Answer:
SAS libraries: click on the Explorer tab and then on Libraries
What is a library? A library is a folder that

contains special SAS files including SAS
data sets. Three libraries (at least), exist
when you start SAS. Other libraries may be
added by you when you create permanent
SAS data sets.
Default SAS Libraries:
SASHELP contains SAS help files
SASUSER contains device drivers, fonts

and the user profiles
WORK will contain temporary SAS

data sets
Here we also see a library called MAPS.

BIO113 28
Temporary vs. permanent SAS data sets:
Temporary SAS file: Exists only for the duration of the SAS session
Is stored in the library, WORK (click on WORK to see what's
been stored there)
You gave it a 'single-level' name (though SAS gives it a 'two-
level' name
Permanent SAS file: Exists after you exit a SAS session (we'll make one shortly!)
Is stored in your own library
You gave it a 'two-level' name
Create a temporary SAS data set: (we've been doing it)
Example:
* creating a temporary SAS dataset; Data
filename pups 'p:\bio113\ex22.dat';
data dogs; spot 2
infile pups; dot 4
input name $ age;
run;
LOG: 15 * creating a temporary SAS dataset;

16 filename pups 'p:\bio113\ex22.dat';
17 data dogs;
18 infile pups;
19 input name $ age;
20 run;
NOTE: The infile PUPS is:

File Name=P:\bio113\ex22.dat,
RECFM=V,LRECL=256
NOTE: 2 records were read from the infile PUPS.

NOTE: The data set WORK.DOGS has 2 observations and 2
variables.
Result: Examine the SAS log. SAS created DOGS with 2 observations and 2 variables.
While you gave the data set the single-level name, DOGS, SAS knows it by a
two-level name, WORK.DOGS. The first level of the name WORK) is the libref
and it points to the library (physically, a subfolder in c:\SASTemp\_Txxxx; xxxx is
a number generated for this session) where temporary SAS data files are stored.
The second level is DOGS, the name you gave the data set. Click Libraries in
Explorer and the library WORK to verify that DOGS is there.
BIO113 29
Test: Write a program to create a temporary data set with the single-level name,
INSECTS, and with variables, BUG and LEGS. Then see if you can find it in a
SAS library. Which library should you look in?
Here are the data lines in ex23.dat:
spider 8
cricket 6
Answer:
BIO113 30
Create a permanent SAS data set:
How? Tell SAS the name of the library (folder) where you want to store the SAS data set.
LIBNAME libref 'library-name';
Example: * creating a permanent SAS data set (libref is muffy);

libname muffy 'p:\bio113';
data muffy.dogs;
input name $ age;
run;
LOG: 17 * creating a permanent SAS dataset (libref is muffy);

18 libname muffy 'p:\bio113';
NOTE: Libref MUFFY was successfully assigned as follows:
Engine: V8
Physical Name: p:\bio113
19 data muffy.dogs;
20 infile 'p:\bio113\ex22.dat';
22 run;
NOTE: The infile 'p\bio113\ex22.dat' is:

RECFM=V,LRECL=256

NOTE: The data set MUFFY.DOGS has 2 observations and 2 variables.
Result: Examine the SAS Log. SAS created MUFFY.DOGS with 2 observations and 2
variables. The first level of the name (muffy) is the libref I chose and it points
to a library that's physically a folder (p:\bio113) I created before I ran the
program. This permanent SAS data set will be stored in p:\bio113. Click
Libraries and under Active Libraries, muffy, to verify; notice that SAS calls the
file DOGS (Not muffy.dogs). Just to confuse things, the file will have the name
dogs.sas7bdat (sas7bdat is the special suffix reserved for version 7 and 8
SAS data sets) when you look for it with Windows “My Computer” or “Search”.
Test: Write a program to read the insect data set (ex23.dat), as before, but, this time,
save it as a permanent SAS data set, insects.sas7bdat, on your P drive.
Answer:
BIO113 31
Using a permanently stored SAS data set:
Example: Come back later to read the permanent data set, dogs.sas7bdat, into a
temporary data set (foo) and print the data. Note, this time the libref is alvin
but it points to the same physical location, p:\bio113.
Program: * read from an existing SAS data set;

libname alvin 'p:\bio113';
data foo;
set alvin.dogs; * set is the SAS command that reads existing SAS data sets *;
run;
proc print; run;
LOG: 23 * read from an existing SAS data set;

24 libname alvin 'p:\bio113';
NOTE: Libname ALVIN refers to the same physical library as
MUFFY.
NOTE: Libref ALVIN was successfully assigned as follows:
Engine: V8
Physical Name: p:\bio113
25 data foo;
26 set alvin.dogs; * set is a SAS command that reads
existing SAS data sets *;
27 run;
NOTE: There were 2 observations read from the dataset

ALVIN.DOGS.
NOTE: The data set WORK.FOO has 2 observations and 2
variables.
Test: Write SAS statements to read the insect data set (ex23.dat) and save
insects.sas7bdat permanently in your own folder, p:\bio113. Write statements
that read insects.sas7bdat from your folder into a temporary data set (call it
ONE) and then print the data.
Answer:
BIO113 32
Another way to create and later use a permanent SAS data set:
Starting with SAS version 7, you can directly reference SAS data sets in programs. The
syntax is similar to that of the INFILE statement where the name of a raw data set is
included in quotation marks. On the DATA (or SET, MERGE, etc) statement, include the
path and member name of the SAS data set in quotes. Notice that SAS, in fact, creates a
libref for you (Wc000001, 2, etc) and lists it under Active Libraries. In this program the SAS
file created and later read is dogs.sas7bdat and it can be found in the folder, p:\bio113.
Example: * create a permanent SAS data set without a libref;

data 'p:\bio113\dogs';
input name $ age;
run;
LOG: 15 data 'p:\bio113\dogs';

16 infile 'p:\bio113\ex22.dat';
18 run;
NOTE: The infile 'p:\bio113\ex22.dat' is:

RECFM=V,LRECL=256

NOTE: The data set p:\bio113\dogs has 2 observations and 2 variables.
Test: Write SAS statements to read the insect data set (ex23.dat) and save
insects.sas7bdat permanently in your own folder, p:\bio113 without using a
libref statement.
Answer:
BIO113 33
Read a permanent SAS data set without using a libref statement:
Example: * read a permanent SAS data set without a libref;

data foo;
set 'p:\bio113\dogs';
run;
LOG: 21 data foo;

22 set 'p:\bio113\dogs';
23 run;
NOTE: There were 2 observations read from the dataset

p:\bio113\dogs.
NOTE: The data set WORK.FOO has 2 observations and 2 variables.
cpu time 0.03 seconds.
Test: Read your permanent SAS data set, insects.sas7bdat, into a temporary data set, foo.
Answer:
BIO113 34
See what's in the SAS data set:
PROC CONTENTS: This PROC lists documentation stored with a temporary or permanent
SAS data set.
Example: * what's in the SAS data set?;

libname alvin 'p:\bio113';
proc contents data=alvin.dogs;
run;
/* here is another proc contents statement */

/* it's commented out but we can try it in class */
/* proc contents position data=alvin.dogs; run; */
Result:
The CONTENTS Procedure
Data Set Name: ALVIN.DOGS Observations: 2

Member Type: DATA Variables: 2
Engine: V8 Indexes: 0
Created: 16:07 Thursday, Observation Length: 16
August 26, 2004
Last Modified: 16:07 Thursday, Deleted Observations: 0
August 26, 2004
Protection: Compressed: NO
Data Set Type: Sorted: NO
Label:
-----Engine/Host Dependent Information-----
Data Set Page Size: 4096

Number of Data Set Pages: 1
First Data Page: 1
Max Obs per Page: 252
Obs in First Data Page: 2
Number of Data Set Repairs: 0
File Name: p:\bio113\dogs.sas7bdat
Release Created: 8.0202M0
Host Created: WIN_PRO
-----Alphabetic List of Variables and Attributes-----
# Variable Type Len Pos

-----------------------------------
2 age Num 8 0
1 name Char 8 8
BIO113 35
View the data: You may also view the data contained in the SAS data set. We've been
doing that using PROC PRINT but it can also be done from SAS's Explorer. Click Explorer,
Libraries and then click the active library WORK. Then double click Foo. (Note that if you
right click on the file and choose View Columns from the pull-down menu, you get an
abbreviated 'proc contents'.)
Here's what you see.
Here’s what you see with ‘View Columns'.
Click on the u to the right of Columns to choose to see the data set properties or engine
and host information.
BIO113 36
INPUT FROM OTHER PROGRAMS: STAT/TRANSFER™
Stat/Transfer is a separate, easy to use, Windows program that transfers data among
several statistical, database and spreadsheet programs. Among the input and output files
Stat/Transfer works with are ASCII, Access, dBase Excel, Lotus 1-2-3, Paradox, SAS, S-
plus, SPSS, Stata and Systat.
Start Stat/Transfer by clicking Start, Miscellaneous and StatTransfer. You will see this
dialogue box.
Select an input file type—I've indicated a Stata data set—and click Browse to locate the file.
My Stata file is c:\bio113\endomet.dta. The suffix, .dta, indicates this is a Stata data set.
Next, select the output file type—I've indicated a SAS for Windows file. Stat/Transfer gave
the output file the same name as the input file except the suffix is .sas7bdat, which indicates
a Version 7/8 SAS data set is to be created. You may change any part of this name (use
Browse) but do not change the suffix. Finally, click Transfer and the SAS data set will be
created from the Stata data set. The tabs--Variables, Observations and Options--permit you
to transfer subsets of variables and observations and to select various transfer options.
When you create a SAS data set with Stat/Transfer you avoid the INPUT issue entirely! You
may immediately access the data set with SAS's SET statement. We will demonstrate this
handy program in class.
BIO113 37
CREATE NEW VARIABLES & CHANGE EXISTING VARIABLES (LSB 3.1-3.3, 3.7-3.8, 8.7-8.9,
8.12)
Assignment statements: New variables are created and existing ones modified with
assignment statements. Note that these SAS statements do not
begin with SAS keywords.
Create new variables: Assign a constant value or the result of an expression to a new
variable. The new variable appears to the left of the equals sign.
If any variable used in an expression is missing (.), the new
variable will also be missing.
Example program: Values of variables in foo1.sas7bdat
libname in 'p:\bio113'; id temp a b c

data one; set in.foo1;
1 98.6 1 . 3
name='Molly'; * note quotes on char var; 2 99.1 4 5 6
age=5.2; * note NO quotes on num var;
age_mons=age*12;
tempc=5/9*(temp-32);
x=12.5+7.8+2.4;
y=a+b+c;
run;
proc print; run;
Result:
Obs id temp a b c name age age_mons tempc x y
1 1 98.6 1 . 3 Molly 5.2 62.4 37.0000 22.7 .

2 2 99.1 4 5 6 Molly 5.2 62.4 37.2778 22.7 15
BIO113 38
Change existing variables: Replace the value of a variable with a new value.
libname in 'p:\bio113'; id age temp a b c

1 5.2 100 1 2 3
age=age*12; 2 3.6 37 5 2 .
temp=9/5*temp+32;
avg1=(a+b+c)/3;
a=a+b+c;
avg2=(a+b+c)/3;
run;
proc print; run;
Result:
Obs id age temp a b c avg1 avg2
1 1 62.4 212.0 6 2 3 2 3.66667

2 2 43.2 98.6 . 2 . . .
BIO113 39
SAS functions: SAS has a library of functions that perform arithmetic, string and other
operations on constants and variables. Use these functions in expressions to
create new variables or replace existing ones. Unlike the assignment
statements above, missing values are not always generated by missing data
when you use SAS functions. We will find a listing of built-in SAS functions
using SAS on-line Help (Click in succession: Help, Getting Started with SAS
Software, Help on SAS software products, Base software, Using base SAS
software, Working with the SAS language, SAS Functions, Function
Categories).
libname in 'p:\bio113'; id bp1 bp2 bp3 first last

1 116 122 129 yogi bear
xbp=mean(101,107,118); 2 111 . 114 rocky squirrel
ybp=mean(bp1,bp2,bp3); 3 . . . micky mouse
zbp=mean(of bp1-bp3);
name=trim(last)||', '||trim(first);
init=substr(first,1,1);
tot1=round(sum(2.2,1.3,6.8));
tot2=round((sum(2.2,1.3,6.8)),.5);
true=sum(of bp1-bp3) > 350;

run;
proc print; run;
Result:
Obs id bp1 bp2 bp3 first last xbp ybp
1 1 116 122 129 yogi bear 108.667 122.333

2 2 111 . 114 rocky squirrel 108.667 112.500
3 3 . . . micky mouse 108.667 .
Obs zbp name init tot1 tot2 true
1 122.333 bear, yogi y 10 10.5 1

2 112.500 squirrel, rocky r 10 10.5 0
3 . mouse, micky m 10 10.5 0
BIO113 40
Date functions: SAS performs date/time arithmetic on dates expressed as number of days
before or since January 1, 1960 (1/1/1960 has the value 0) and time
expressed as seconds since midnight (midnight has the value 0). Date-time is
expressed as seconds before or since midnight, January 1, 1960. Earlier, we
used the date informat mmddyy8. to input a date, 01161960, which was then
represented in the SAS data set as 15. SAS date and time functions create
SAS dates from variables containing values for month, day, year, hours,
minutes, seconds, etc, or from constants. More date/time functions are listed
in SAS on-line help.
libname in 'p:\bio113'; id bm bd by bh bmn

1 1 16 1960 2 10
dob1=mdy(1,16,1960); 2 12 25 1959 3 22
tob1=hms(2,10,0);
dob2=mdy(bm,bd,by);
tob2=hms(bh,bmn,0);
dob3=dhms(dob2,bh,bmn,0);
thisday=today();
now=time();
rightnow=24*60*60*thisday+now;
run;
proc print; run;
Result:
The SAS System 13:37 Wednesday, September 1, 2004
Obs id bm bd by bh bmn dob1 tob1 dob2 tob2 dob3 thisday now rightnow
1 1 1 16 1960 2 10 15 7800 15 7800 1303800 16315 49296.16 1409665296.2

2 2 12 25 1959 3 22 15 7800 -7 12120 -592680 16315 49296.16 1409665296.2
BIO113 41
Year 2K?
yearcutoff=nnnn where nnnn is the first year of the 100-year span. The yearcutoff option
applies only to dates where year is described with 2 digits. Notice the effect of the two
output formats, mmddyy8. and mmddyy10.
The default value for yearcutoff is 1920.
Program: Data:
options nocenter yearcutoff=1940; 1 041500

filename y2k 'p:\bio113\ex24.dat'; 2 041519
data one; infile y2k; 3 041520
input id dob mmddyy6.; 4 041539
dob1=dob; 5 041540
interval=(today()-dob)/365.25; 6 041559
7 041560
run; 8 041599
proc print; id id;
var dob dob1 interval;
format dob mmddyy8. dob1 mmddyy10.;
title 'Dates with yearcutoff=1940';
run;
Result:
Dates with yearcutoff=1940 15:15 Sunday, August 15, 2004
id dob dob1 interval
1 04/15/00 04/15/2000 4.3340

2 04/15/19 04/15/2019 -14.6639
3 04/15/20 04/15/2020 -15.6660
4 04/15/39 04/15/2039 -34.6639
5 04/15/40 04/15/1940 64.3340
6 04/15/59 04/15/1959 45.3361
7 04/15/60 04/15/1960 44.3340
8 04/15/99 04/15/1999 5.3361
This is the output when you omit 'yearcutoff=1940' from the program:
Dates with yearcutoff=1920 15:15 Sunday, August 15, 2004
id dob dob1 interval
1 04/15/00 04/15/2000 4.3340

2 04/15/19 04/15/2019 -14.6639
3 04/15/20 04/15/1920 84.3340
4 04/15/39 04/15/1939 65.3361
5 04/15/40 04/15/1940 64.3340
6 04/15/59 04/15/1959 45.3361
7 04/15/60 04/15/1960 44.3340
8 04/15/99 04/15/1999 5.3361
BIO113 42
Date difficulties
Occasionally you'll be given a SAS data set with dates that were read in without using SAS
date functions. The variables may be stored as character or numeric but SAS will not have
interpreted them as dates. You will need to disassemble the variables into months, days and
years and then apply a SAS date function to get a proper SAS date. See if you can figure out
how to make SAS times out of time information that was read in incorrectly.
* extract month, day, and year so you can use the mdy function when a date
is read from raw data as a character or numeric variable instead of
as a date variable;
data one;
input cdate $10. +1 ndate 8.;
mon1=1*substr(cdate,1,2);
day1=1*substr(cdate,4,2);
yr1 =1*substr(cdate,7,4);
mon2=int(ndate/1000000);
day2=int(ndate/10000)-(mon2*100);
yr2 =ndate-(mon2*1000000+day2*10000);
date1=mdy(mon1,day1,yr1);
/* or date2=mdy(mon2,day2,yr2); */
datalines;
09-30-2003 09302003
11/06/1970 11061970
run;
proc print; run;
The SAS System
Obs cdate ndate mon1 day1 yr1 mon2 day2 yr2 date1
1 09-30-2003 9302003 9 30 2003 9 30 2003 15978

2 11/06/1970 11061970 11 6 1970 11 6 1970 3962
BIO113 43
Useful SAS functions
These SAS functions were used in a program executed on 09/01/2004 at 1:37:03 PM.
Data:
x1 x2 x3 x4 x5 f nm a1 a2 mon day yr hr min sec
10 5 23 . 37 21.655 -4 Muffy st♦♦♦ 9 01 2004 13 37 3
Function SAS assignment statement Answer (the value of 'var')

mean var=mean(3,4,5); 4
var=mean(x1,x2,x3,x4,x5); 18.75
var=mean(of x1-x5); 18.75
sum var=sum(3,4,5); 12
var=sum(x1,x2,x3,x4,x5); 75
var=sum(of x1-x5); 75
min var=min(3,4,5); 3
var=min(x1,x2,x3,x4,x5); 5
var=min(of x1-x5); 5
max var=max(3,4,5); 5
var=max(x1,x2,x3,x4,x5); 37
var=max(of x1-x5); 37
n var=n(x1,x2,x3,x4,x5); 4
var=n(of x1-x5); 4
nmiss var=nmiss(x1,x2,x3,x4,x5); 1
var=nmiss(of x1-x5); 1
round var=round(11.336); 11
var=round(f,.01); 21.66
abs var=abs(nm); 4
int var=int(f); 21
mod var=mod(f,1); 0.655
var=mod(x1,3); 1
exp var=exp(x2); 148.413
log var=log(x1); 2.30259
log10 var=log10(x1); 1
sqrt var=sqrt(x1); 3.16228
** var=x2**3; 125
mdy date=mdy(9,30,2003); 16315 (01SEP2004)
date=mdy(mon,day,yr); 16315 (01SEP2004)
hms var=hms(hr,min,sec); 49023 (13:37:03)
dhms var=dhms(date,hr,min,sec); 1409665023 (01SEP2004:13:37:03)
today var=today(); 16315 (01SEP2004)
time var=time(); 49023 (13:37:03)
substr var=substr(a1,3,2); ff
trim var=trim(a2); st
var=trim(a2)||a1; stMuffy
upcase var=upcase(a1); MUFFY
lowcase var=lowcase(a1); muffy
BIO113 44
BIO113 45
CONDITIONAL PROCESSING (LSB 3.4-3.6)
Conditional assignment: IF-THEN, IF-THEN-DO-END
IF-THEN permits processing of subsets of observations. Only a single assignment or action

may be made after THEN.
Program: Values of variables in foo5.sas7bdat
libname in 'p:\bio113'; name age sex wt x1 x2 x3 x4

molly 9 m 30 1 . 3 3
if name='molly' then sex='f'; max 12 m 54 5 3 2 7
mandy 99 f . 1 1 1 1
if age < 10 then agecat1=1; mike . m 32 . 5 . 7
matt 0 62 3 4 1 6
* if . < age < 10 then agecat2=1;
if 30 <= wt < 40 and sex='m' then group=1;
if age=0 or age=99 then age=.;
if n(of x1-x4) = 4 then complete=1;
miss=nmiss(of x1-x4) > 0;
if name='mandy' then delete;

run;
proc print; run;
Result:
name age sex wt x1 x2 x3 x4 agecat1 agecat2 group complete miss
molly 9 f 30 1 . 3 3 1 1 . . 1
max 12 m 54 5 3 2 7 . . . 1 0
mike . m 32 . 5 . 7 1 . 1 . 1
matt . 62 3 4 1 6 1 1 . 1 0
* NOTE:
the statement: if . < age <= 10 then agecat2=1;
could be written: if age > . & age <= 10 then agecat2=1;
or: if age ^= . & age <= 10 then agecat2=1;

BIO113 46
IF-THEN-DO-END: If you want to make more than one assignment, the syntax changes slightly.
IF condition THEN DO; Values of variables in foo5.sas7bdat

assignment 1
assignment 2 ... name age sex wt x1 x2 x3 x4
END; molly 9 m 30 1 . 3 3
max 12 m 54 5 3 2 7
mandy 99 f . 1 1 1 1
Program: mike . m 32 . 5 . 7
matt 0 62 3 4 1 6
libname in 'p:\bio113';
if name='molly' | . < age < 10 then do;

sex='f';
agecat=1;
end;
if nmiss(of x1-x4)=0 then do;

complete=1;
ratio=wt/age;
end;
run;
proc print; id name;

run;
Result:
Obs name age sex wt x1 x2 x3 x4 agecat complete ratio
1 molly 9 f 30 1 . 3 3 1 . .
2 max 12 m 54 5 3 2 7 . 1 4.5
3 mandy 99 f . 1 1 1 1 . 1 .
4 mike . m 32 . 5 . 7 . . .
5 matt 0 f 62 3 4 1 6 1 1 .
BIO113 47
Multiple choice processing: IF-THEN-ELSE
If the condition is not met, what value is assigned to the variable?

Unless you tell SAS otherwise, the value is missing. Use IF-THEN-ELSE to make different
assignments when different conditions are met. Multiple assignments are made using
DO...END as before.
Example program Values of variables in foo6.sas7bdat
libname in 'p:\bio113'; id name age bp

1 ham 52 100
if . < age <= 10 then agecat=1; 2 pam 20 70
else if 10 < age <= 20 then agecat=2; 3 nan . 80
else if age > 20 then agecat=3; 4 sam 8 78
4 dan 18 .
if . < bp <= 90 then htn1=0; 6 ann 26 .
else if bp > 90 then htn1=1;
htn2=(bp > 90);
if . < age <= 20 then do;

if . < bp <= 80 then agehtn=0;
else agehtn=1;
end;
else if age > 20 then do;
if . < bp <= 90 then agehtn=0;
else agehtn=1;
end;
if name='pam' | name='ann' then do;

sex='f';
group=5;
end;
else if name='sam' then do;
sex='m';
group=3;
end;
run;
proc print;
run;
Result:
Obs id name age bp agecat htn1 htn2 agehtn sex group
1 1 ham 52 100 3 1 1 1 .
2 2 pam 20 70 2 0 0 0 f 5
3 3 nan . 80 . 0 0 . .
4 4 sam 8 78 1 0 0 0 m 3
5 4 dan 18 . 2 . 0 1 .
6 6 ann 26 . 3 . 0 1 f 5
BIO113 48
Subsetting IF: Send a subset of observations to the next DATA step
A subsetting IF statement of the form IF condition; is equivalent to these statements:
IF condition then OUTPUT;

IF not-condition THEN DELETE;
Obs id name age bp agecat htn1 htn2 agehtn sex group
1 1 ham 52 100 3 1 1 1 .
2 2 pam 20 70 2 0 0 0 f 5
3 3 nan . 80 . 0 0 . .
4 4 sam 8 78 1 0 0 0 m 3
5 4 dan 18 . 2 . 0 1 .
6 6 ann 26 . 3 . 0 1 f 5
Example program (make subsets of WORK.ONE, above, from previous page)
Result:
data two; set one; Obs name age sex

if sex ^= 'f';
* if sex ^= 'f' then output; 1 ham 52
* if sex = 'f' then delete; 2 nan .
run; 3 sam 8 m
proc print; 4 dan 18
var name age sex;
run;

if age < 20 & sex='m';
run; 1 sam 8 m
proc print;
var name age sex;
run;

if age < 20 | sex='m';
run; 1 nan .
proc print; 2 sam 8 m
var name age sex; 3 dan 18
run;
/* watch out on this one! */ Obs name age sex

data two; set one;
if age < 20; 1 sam 8 m
if sex='m';
run;
proc print;
var name age sex;
run;
BIO113 49
ARRAYS (LSB 3.10)
Previously, we processed subsets of observations. ARRAY permits processing of subsets of

variables. All variables in a particular array must be of the same data type. First you define
the array and then you may process it.
Define arrays: It is my habit to use caps for array names--it makes them stand out from the
variables! In any event, arrays are named using the rules for all SAS names.
In an explicit array, the number of variables in the array or '*' appears in
parentheses after the array name. In an implicit array, the number in
parentheses is omitted. Arrays may be multidimensional but, if so, must be
explicit. A $ indicates that the array contains character variables.
Example: array MISS(3) ga wt len;

array SS(6) score1-score6;
array SX(*) $ neuro cv gi ent;
array MULTI(3,2) a1 a2 b1 b2 c1 c2;
array MISS weight--age;
Process arrays: Arrays are usually processed in DO...END loops. In an explicit array, the
number of times the loop is processed (usually it's the number of variables in
the array) is specified by a subscript variable. You may process all variables
in an array in the same way or you may address a particular variable by
number (subscript). The first variable in the array has subscript, 1, the second,
2, and so forth. Several arrays may be processed in one DO...END if they
contain the same number of variables.
Example:
libname in 'p:\bio113'; id ga wt len

* drop i; 1 29 1150 9999
array MISS(3) ga wt len; 2 9999 764 18
do i=1 to 3;
if MISS(i)=9999 then MISS(i)=.;
end;
run;
proc print;
run;
Result:
Obs id ga wt len i
1 1 29 1150 . 4
2 2 . 764 18 4
BIO113 50
* again with put _all_; id ga wt len

data one; set in.foo7; 1 29 1150 9999
* drop i; 2 9999 764 18
array MISS(3) ga wt len;
do i=1 to 3;
put 'before: ' _all_;
if MISS(i)=9999 then MISS(i)=.;
put 'after: ' _all_;
end;
run;
proc print;
run;
The output is identical but look at the SAS log to see how values are reassigned by the
array.
213 * again with put _all_;

214 libname in 'c:\bio113\data';
NOTE: Libref IN was successfully assigned as follows:
Engine: V8
Physical Name: c:\bio113\data
215 data one; set in.foo7;
216 * drop i;
217 array MISS(3) ga wt len;
218 do i=1 to 3;
219 put 'before: ' _all_;
220 if MISS(i)=9999 then MISS(i)=.;
221 put 'after: ' _all_;
222 end;
223 run;
before: id=1 ga=29 wt=1150 len=9999 i=1 _ERROR_=0 _N_=1

after: id=1 ga=29 wt=1150 len=9999 i=1 _ERROR_=0 _N_=1
after: id=1 ga=29 wt=1150 len=9999 i=2 _ERROR_=0 _N_=1
after: id=1 ga=29 wt=1150 len=. i=3 _ERROR_=0 _N_=1
after: id=2 ga=. wt=764 len=18 i=1 _ERROR_=0 _N_=2
before: id=2 ga=. wt=764 len=18 i=2 _ERROR_=0 _N_=2
before: id=2 ga=. wt=764 len=18 i=3 _ERROR_=0 _N_=2
BIO113 51
Example:
libname in 'p:\bio113'; id resp cv gi heme

* drop j; 1 n y y n
array SX(*) $ resp cv gi heme; 2 n n n n
count=0; 3 y n y y
do j=1 to dim(SX);
if SX(j)='y' then count=count+1;
end;
run;
proc print;
run;
Result:
Obs id resp cv gi heme count j
1 1 n y y n 2 5
2 2 n n n n 0 5
3 3 y n y y 3 5
libname in 'p:\bio113'; id sbp1 sbp2 sbp3 dbp1 dbp2 dbp3

* drop rep; 1 120 115 132 87 70 79
array SS(3) sbp1 sbp2 sbp3; 2 119 123 127 78 75 86
array DD(3) dbp1 dbp2 dbp3;
array RR(3) rat1 rat2 rat3;
do rep=1 to 3;
RR(rep)=SS(rep)/DD(rep);
end;
run;
proc print;
run;
Result:
Obs id sbp1 sbp2 sbp3 dbp1 dbp2 dbp3 rat1 rat2 rat3 rep
1 1 120 115 132 87 70 79 1.37931 1.64286 1.67089 4

2 2 119 123 127 78 75 86 1.52564 1.64000 1.47674 4
BIO113 52

* drop rep; 1 120 115 132 87 70 79
array BB(2,3) sbp1-sbp3 dbp1-dbp3; 2 119 123 127 78 75 86
array RR(3) rat1-rat3;
do rep=1 to 3;
RR(rep)=BB(1,rep)/BB(2,rep);
put _all_;
end;
run;
proc print;
run;
Result:
139 libname in '.';
NOTE: Libref IN was successfully assigned as follows:
Engine: V8
Physical Name: P:BIO113
140 data one; set in.foo9;
141 * drop rep;
142 array BB(2,3) sbp1-sbp3 dbp1-dbp3;
143 array RR(3) rat1-rat3;
144 do rep=1 to 3;
145 RR(rep)=BB(1,rep)/BB(2,rep);
146 put _all_;
147 end;
148 run;
id=1 sbp1=120 sbp2=115 sbp3=132 dbp1=87 dbp2=70 dbp3=79 rat1=1.3793103448 rat2=. rat3=.
rep=1_ERROR_=0 _N_=1
id=1 sbp1=120 sbp2=115 sbp3=132 dbp1=87 dbp2=70 dbp3=79 rat1=1.3793103448
rat2=1.6428571429 rat3=. rep=2 _ERROR_=0 _N_=1
id=1 sbp1=120 sbp2=115 sbp3=132 dbp1=87 dbp2=70 dbp3=79 rat1=1.3793103448
rat2=1.6428571429 rat3=1.6708860759 rep=3 _ERROR_=0 _N_=1
id=2 sbp1=119 sbp2=123 sbp3=127 dbp1=78 dbp2=75 dbp3=86 rat1=1.5256410256 rat2=. rat3=.
rep=1 _ERROR_=0 _N_=2
id=2 sbp1=119 sbp2=123 sbp3=127 dbp1=78 dbp2=75 dbp3=86 rat1=1.5256410256 rat2=1.64
rat3=. rep=2 _ERROR_=0 _N_=2
id=2 sbp1=119 sbp2=123 sbp3=127 dbp1=78 dbp2=75 dbp3=86 rat1=1.5256410256 rat2=1.64
rat3=1.476744186 rep=3 _ERROR_=0 _N_=2
NOTE: There were 2 observations read from the dataset IN.FOO9.
149 proc print;

150 run;
NOTE: There were 2 observations read from the dataset WORK.ONE.

NOTE: PROCEDURE PRINT used:
Obs id sbp1 sbp2 sbp3 dbp1 dbp2 dbp3 rat1 rat2 rat3 rep
1 1 120 115 132 87 70 79 1.37931 1.64286 1.67089 4

2 2 119 123 127 78 75 86 1.52564 1.64000 1.47674 4
BIO113 53
libname in 'p:bio113'; id a b c d e
* notice NO drop statement; 1 229 9999 871 381 173
array MISS a--e; 2 9999 637 377 9999 383
do over MISS;
if MISS=9999 then MISS=.;
end;
run;
proc print; run;
Result:
Obs id a b c d e
1 1 229 . 871 381 173

2 2 . 637 377 . 383

* notice NO drop statement; 1 120 115 132 87 70 79
array SS sbp1 sbp2 sbp3; 2 119 123 127 78 75 86
array DD dbp1 dbp2 dbp3;
array RR rat1 rat2 rat3;
do over RR; * or SS or DD;
RR=SS/DD;
end;
run;
proc print;
run;
Result:
Obs id sbp1 sbp2 sbp3 dbp1 dbp2 dbp3 rat1 rat2 rat3
1 1 120 115 132 87 70 79 1.37931 1.64286 1.67089

2 2 119 123 127 78 75 86 1.52564 1.64000 1.47674
BIO113 54
Create observations: Up to now, we've read in a single observation and placed a single
observation in a SAS data set or we've deleted it (subsetting IF). An OUTPUT statement inside
the DO...END loop that processes an array, will cause an observation to be output each time the
array is processed so that multiple observations are placed in a new SAS data set. This program
changes the unit of observation in the data set from a subject to a subject-visit.
Example:
libname in 'p:\bio113'; id wt1 wt2 wt3

1 88 90 97
array WW(3) wt1-wt3; 2 91 85 87
do visit=1 to 3;
wt=WW(visit);
output;
end;
run;
proc print;
run;
Result:
Obs id wt1 wt2 wt3 visit wt
1 1 88 90 97 1 88
2 1 88 90 97 2 90
3 1 88 90 97 3 97
4 2 91 85 87 1 91
5 2 91 85 87 2 85
6 2 91 85 87 3 87
Program:
drop wt1-wt3; * but don't drop visit;
array WW(3) wt1-wt3;
do visit=1 to 3;
wt=WW(visit);
output;
end;
run;
proc print;
run;
Result:
Obs id visit wt
1 1 1 88
2 1 2 90
3 1 3 97
4 2 1 91
5 2 2 85
6 2 3 87
BIO113 55
Program:
drop wt1-wt3;
array WW wt1-wt3;
do over WW;
wt=WW;
visit=_i_;
output;
end;
run;
proc print;
run;
Result:
Obs id visit wt
1 1 1 88
2 1 2 90
3 1 3 97
4 2 1 91
5 2 2 85
6 2 3 87
Program:
* SAS log with put _all_;
drop wt1-wt3;
array WW wt1-wt3;
do over WW;
wt=WW;
visit=_i_;
put _all_;
output;
end;
run;
From the SAS log:
id=1 wt1=88 wt2=90 wt3=97 _I_=1 wt=88 visit=1 _ERROR_=0 _N_=1

BIO113 56
SORTING AND PRINTING DATA WITH PROCS: (LSB 4.1-4.7)
Basics: Procedures process the most recently created SAS data set unless you explicitly
name another one
Most procedures produce some sort of a report in the OUTPUT window but SORT
doesn't so we must print to see what sort has done
SORT modifies the data set—it changes the order of observations in the data set
(verify this by looking at the data with VIEW between sorts).
Process a subset (without changing the SAS data set) with WHERE
Example: Create a SAS data set then PROC it (this is one SAS session)
* create a SAS data set and run PROCs; Data:

data in.moredogs; spot 4 1 5
infile 'p:\bio113\ex25.dat'; dot 2 0 3
input name $ age sex ribbons; cleo 1 0 8
run; oreo 1 1 4
PROCedure statements Result:
proc sort data=in.moredogs; Obs name age sex ribbons

by name;
run; 1 cleo 1 0 8
proc print; 2 dot 2 0 3
run; 3 oreo 1 1 4
4 spot 4 1 5
proc sort; by descending ribbons; Obs name age ribbons

run;
proc print; 1 cleo 1 8
var name age ribbons; 2 spot 4 5
run; 3 oreo 1 4
4 dot 2 3
proc sort; by age descending ribbons; Obs name age ribbons

run;
proc print; 1 cleo 1 8
var name age ribbons; 2 oreo 1 4
run; 3 dot 2 3
4 spot 4 5
BIO113 57
PROCedure statements Result:
proc sort; by sex name; sex=0

run;
proc print; by sex; id name; name age ribbons
sum ribbons;
var age ribbons; cleo 1 8
run; dot 2 3
---- -------
sex 11
sex=1
name age ribbons
oreo 1 4
spot 4 5
---- -------
sex 9
=======
20
proc print; where sex=0; Obs name ribbons

var name ribbons;
run; 1 cleo 8
2 dot 3
proc sort nodupkey out=one; Obs name age sex ribbons

by age;
run; 1 cleo 1 0 8
proc print; 2 dot 2 0 3
run; 3 spot 4 1 5
BIO113 58
Dress up the output:
Label: Fuller description of variables

Each label may be up to 256 characters in length including blanks
Must be enclosed in single or double quotes
Remains with variable if stated in the DATA step
Disappears after step if stated in the PROC step
Example: label ribbons='Number of first place ribbons';

label age="Dog's age" sex='Dog''s sex';
Title: Label top of page
Footnote: Label bottom of page
Each of 10 titles and footnotes may be up to 256 characters including blanks

Must be enclosed in matched single or double quotes
Appear anywhere in program and remain until replaced by another
To return to SAS defaults: TITLE;
FOOTNOTE;
Example: title 'Ourtown Dog Show';
title1 'Ourtown Dog Show';

title2 'Winner''s Summary';
footnote 'Stats compiled by judges';
Program (continue with moredogs.sas7bdat. Note: data= is required):

* print in.moredogs (previous page) with titles, footnote, labels;
proc print data=in.moredogs label; id name;
label ribbons='Number of first place ribbons'
age="Dog's age" sex='Dog''s sex';
title1 'Ourtown Dog Show';
title2 'Winner''s Summary';
footnote 'Stats compiled by judges';
run;
Result: Ourtown Dog Show

Winner's Summary
Number of
Dog's Dog's first place
name age sex ribbons
cleo 1 0 8
dot 2 0 3
oreo 1 1 4
spot 4 1 5
Stats compiled by judges

BIO113 59
Format: Label values of variables

Use SAS built-in formats (date, time, dollars, etc)
Create your own with PROC FORMAT (Up to 200 characters and spaces enclosed
in quotes although truncated in many PROCs)
Remains with variable and change how it's stored if stated in the DATA step
Disappears after step if stated in the PROC step
Example: Use SAS formats Data

Program: data one; harry3.565 9jun1996
infile 'p:\bio113\ex26.dat'; mary 10.33122nov1994
input name $5. weight 6. dob date9.;
run;
proc print;
var name weight dob;
format weight 4.1 dob mmddyy10.;
run;
Result:
Obs name weight dob
1 harry 3.6 06/09/1996

2 mary 10.3 11/22/1994
Example: Create your own formats
Format names, which appear after the key word VALUE, may not
exceed 8 characters and may not end with a number.
Program: * print variables in in.moredogs;
proc format;
value sex 0='female' 1='male';
value age 0-1='puppy' 2-high='adult';
run;
proc print data=in.moredogs;
var name sex age;
format sex sex. age age.;
run;
proc print data=in.moredogs;
var name sex age;
format sex sex1. age age3.;
run;
Result: Obs name sex age
1 cleo female puppy

2 dot female adult
3 oreo male puppy
4 spot male adult
Obs name sex age
1 cleo f pup
2 dot f adu
3 oreo m pup
4 spot m adu
BIO113 60
Create your own permanent format library:
If you find yourself using the same formats over and over again, you might want to put them
into your own private format library. The libname statement tells SAS where to put the
format library. The format library will always have the name formats.sas7bcat (the suffix is
sas7bcat, not sas7bdat, which reminds SAS that this special data set contains formats not
data).
Example: Create a permanent format library
Program:
libname mylib 'p:\bio113';

proc format library=mylib;
value sex 0='female' 1='male';
value age 0-1='puppy' 2-high='adult';
run;
Example: Return later and print variables using the format library. When you create the
format library you may choose the libref (I used mylib). However, when you
use a format library, its libref must be library.
Program:
* print variables in in.moredogs;

libname library 'p:\bio113';
data one; set in.moredogs;
run;
proc print;
var name sex age;
format sex sex. age age.;
run;
Result:
Obs name sex age
1 cleo female puppy

2 dot female adult
3 oreo male puppy
4 spot male adult
BIO113 61
SIMPLE DATA SUMMARIZATION: (LSB 4.9-4.11, 4.14, 7.2-7.3)
First, create a SAS data set: Data:

* create a data set for summarization; 1 m31146200
data one; 2 f25 56411
infile 'p:\bio113\ex27.dat'; 3 f27 93101
4 f28100210
input id 1-2 sex $3 ga 4-5 bw 6-9 (cs ivh)(1.); 5 m28106711
6 f32132700
label ga='gestational age' 7 m26 88001
bw='birth weight' 8 f33165000
cs='Cesarian delivery' 9 m29125010
ivh='intraventricular hemorrhage'; 10m26 73601
run; 11f30121200
proc print label; id id;
title 'data set for summarization';
run;
Result:
data set for summarization
gestational birth Cesarian intraventricular

id sex age weight delivery hemorrhage
1 m 31 1462 0 0
2 f 25 564 1 1
3 f 27 931 0 1
4 f 28 1002 1 0
5 m 28 1067 1 1
6 f 32 1327 0 0
7 m 26 880 0 1
8 f 33 1650 0 0
9 m 29 1250 1 0
10 m 26 736 0 1
11 f 30 1212 0 0
BIO113 62
Summarize categorical data - PROC FREQ:
Create univariate, bivariate, multivariate tables

Statistics include Fisher's exact test, Χ2, Mantel-Haenszel Χ2, Likelihood Ratio Χ2
Specify [page-variable] * row-variable * column-variable
Appearance is greatly improved by labels and formats.
Program continues:
proc format;
value yn 0='No' 1='Yes';
run;
proc freq; * NOTE: using work.one *;

tables cs sex ivh;
tables sex*ivh / chisq;
tables sex*cs*ivh / expected;
format ivh cs yn.;
run;
Result:
The FREQ Procedure
Cesarian delivery
Cumulative Cumulative
cs Frequency Percent Frequency Percent
--------------------------------------------------------
No 7 63.64 7 63.64
Yes 4 36.36 11 100.00
sex Frequency Percent Frequency Percent
--------------------------------------------------------
f 6 54.55 6 54.55
m 5 45.45 11 100.00
intraventricular hemorrhage
ivh Frequency Percent Frequency Percent
--------------------------------------------------------
No 6 54.55 6 54.55
Yes 5 45.45 11 100.00
BIO113 63
Table of sex by ivh
sex ivh(intraventricular hemorrhage)
Frequency|
Percent |
Row Pct |
Col Pct |No |Yes | Total
---------+--------+--------+
f | 4 | 2 | 6
| 36.36 | 18.18 | 54.55
| 66.67 | 33.33 |
| 66.67 | 40.00 |
---------+--------+--------+
m | 2 | 3 | 5
| 18.18 | 27.27 | 45.45
| 40.00 | 60.00 |
| 33.33 | 60.00 |
---------+--------+--------+
Total 6 5 11
54.55 45.45 100.00
Statistics for Table of sex by ivh
Statistic DF Value Prob

------------------------------------------------------
Chi-Square 1 0.7822 0.3765
Likelihood Ratio Chi-Square 1 0.7899 0.3741
Continuity Adj. Chi-Square 1 0.0764 0.7823
Mantel-Haenszel Chi-Square 1 0.7111 0.3991
Phi Coefficient 0.2667
Contingency Coefficient 0.2577
Cramer's V 0.2667
WARNING: 100% of the cells have expected counts less

than 5. Chi-Square may not be a valid test.
Fisher's Exact Test

----------------------------------
Cell (1,1) Frequency (F) 4
Left-sided Pr <= F 0.9329
Right-sided Pr >= F 0.3918
Table Probability (P) 0.3247

Two-sided Pr <= P 0.5671
Sample Size = 11
BIO113 64
The FREQ Procedure
Table 1 of cs by ivh
Controlling for sex=f
cs(Cesarian delivery)
ivh(intraventricular hemorrhage)
Frequency|
Expected |
Percent |
Row Pct |
---------+--------+--------+
No | 3 | 1 | 4
| 2.6667 | 1.3333 |
| 50.00 | 16.67 | 66.67
| 75.00 | 25.00 |
| 75.00 | 50.00 |
---------+--------+--------+
Yes | 1 | 1 | 2
| 1.3333 | 0.6667 |
| 16.67 | 16.67 | 33.33
| 50.00 | 50.00 |
| 25.00 | 50.00 |
---------+--------+--------+
Total 4 2 6
66.67 33.33 100.00
The FREQ Procedure
Table 2 of cs by ivh
Controlling for sex=m
cs(Cesarian delivery)
ivh(intraventricular hemorrhage)
Frequency|
Expected |
Percent |
Row Pct |
---------+--------+--------+
No | 1 | 2 | 3
| 1.2 | 1.8 |
| 20.00 | 40.00 | 60.00
| 33.33 | 66.67 |
| 50.00 | 66.67 |
---------+--------+--------+
Yes | 1 | 1 | 2
| 0.8 | 1.2 |
| 20.00 | 20.00 | 40.00
| 50.00 | 50.00 |
| 50.00 | 33.33 |
---------+--------+--------+
Total 2 3 5
40.00 60.00 100.00
BIO113 65
Summarize continuous data - PROC MEANS:
Calculates mean, standard deviation, variance, etc. for one or more variables
Use with BY-variable if data set is sorted by BY-variable
Perform paired t-test
Summaries may be output to a new SAS data set
Program:
proc means maxdec=1; *NOTE: still using work.one *;

var ga bw;
run;
Result:
Variable Label N Mean Std Dev Minimum

----------------------------------------------------------------------------
ga gestational age 11 28.6 2.6 25.0
bw birth weight 11 1098.3 320.6 564.0
----------------------------------------------------------------------------
Variable Label Maximum
-----------------------------------------
ga gestational age 33.0
bw birth weight 1650.0
-----------------------------------------
BIO113 66
Program:
proc sort; by ivh;
run;
proc means maxdec=1; by ivh; * with by variable *;
var ga bw;
format ivh yn.;
run;
Result:
intraventricular hemorrhage=No
----------------------------------------------------------------------------
bw birth weight 6 1317.2 222.1 1002.0
----------------------------------------------------------------------------
-----------------------------------------
-----------------------------------------
intraventricular hemorrhage=Yes
----------------------------------------------------------------------------
bw birth weight 5 835.6 192.6 564.0
----------------------------------------------------------------------------
-----------------------------------------
-----------------------------------------
BIO113 67
Program:
proc means maxdec=1; class ivh; * with class variable (no sort) *;
var ga bw;
format ivh yn.;
run;
Result:
intraventricular N
hemorrhage Obs Variable Label N Mean Std Dev
--------------------------------------------------------------------------------
No 6 ga gestational age 6 30.5 1.9
bw birth weight 6 1317.2 222.1
Yes 5 ga gestational age 5 26.4 1.1

bw birth weight 5 835.6 192.6
--------------------------------------------------------------------------------
intraventricular N
hemorrhage Obs Variable Label Minimum Maximum
----------------------------------------------------------------------------
No 6 ga gestational age 28.0 33.0
bw birth weight 1002.0 1650.0
Yes 5 ga gestational age 25.0 28.0

bw birth weight 564.0 1067.0
----------------------------------------------------------------------------
BIO113 68
Program:
* with output of specific statistics to a new data set (two);
proc means noprint mean std; by ivh;
var ga bw;
output out=two
mean=ga bw std=ga_s bw_s;
run;
proc print data=two;
run;
Result:
Obs ivh _TYPE_ _FREQ_ ga bw ga_s bw_s
1 0 0 6 30.5 1317.17 1.87083 222.150

2 1 0 5 26.4 835.60 1.14018 192.552
Program:
proc means noprint mean std data=one; class ivh;

var ga bw;
output out=two
mean=ga bw std=ga_s bw_s;
run;
proc print data=two;
run;
Result:
Obs ivh _TYPE_ _FREQ_ ga bw ga_s bw_s
1 . 0 11 28.6364 1098.27 2.61812 320.551

2 0 1 6 30.5000 1317.17 1.87083 222.150
3 1 1 5 26.4000 835.60 1.14018 192.552
BIO113 69
Display relationships between variables - PROC PLOT:
Produce Y by X scatter plots

Specify plotting symbol(s)
Specify axis range and intervals (tick marks)
Program:
proc plot data=one; * NOTE still using work.one *;

plot bw*ga / haxis=24 to 34 by 2
vaxis=400 to 1800 by 200 vref=1000;
run;
proc plot data=one;
plot bw*ga = sex;
run;
Result: Plot of bw*ga. Legend: A = 1 obs, B = 2 obs, etc.

-+-----------+-----------+-----------+-----------+-----------+-
| |
1800 + +
b | A |
i 1600 + +
r | A |
t 1400 + +
h | A A |
1200 + A +
w | A |
e 1000 +------------------------A------------------------------------+
i | A A |
g 800 + +
h | A |
t 600 + A +
| |
400 + +
| |
-+-----------+-----------+-----------+-----------+-----------+-
24 26 28 30 32 34
gestational age
Plot of bw*ga. Symbol is value of sex.
---+------+------+------+------+------+------+------+------+---
2000 + +
| |
| |
b | |
i | f |
r | |
t 1500 + m +
h | |
| f |
w | m f |
e | |
i | m |
g 1000 + f +
h | m f |
t | |
| m |
| |
| f |
500 + +
---+------+------+------+------+------+------+------+------+---
25 26 27 28 29 30 31 32 33
gestational age
BIO113 70
PROC GPLOT: High resolution plotting in SAS:
GPLOT (graphic plot) makes plots in which points are not limited to locations determined by
dot matrix printers. The plots also can be in color--you can see color on the screen but,
obviously, you need a color printer to get a hard copy. What you get, in spite of being
'graphic', is not publication quality. Notice that SAS opens a fourth window, GRAPH, to
display the results from gplot.
Examples:
data one; set in.ivh1;

run;
* simple plot - gestational age vs birth weight - See top next page for output;
proc gplot;
plot ga*bw;
run;
* HIGHLIGHT SEX;
proc gplot;
plot ga*bw=sex;
run;
* CHANGE SYMBOL COLORS;

proc gplot;
plot ga*bw=sex;
symbol1 color=red;
symbol2 color=blue;
run;
* CHANGE SYMBOL VALUES;

proc gplot;
plot ga*bw=sex;
symbol1 color=red v='f';
symbol2 color=blue v='m';
run;
proc gplot;
plot ga*bw=sex;
symbol1 v=circle;
symbol2 v=plus;
run;
* SPRUCE IT UP SOME MORE;

proc gplot;
plot ga*bw=sex / haxis=axis2 vaxis=axis1;
title 'Plot of Gestational Age versus Birth Weight by Sex';
footnote justify=right 'IVH data set';
axis1 label=('Gestational Age');
axis2 label=('Birth Weight');
run;
* ROTATE THE Y-AXIS LABEL;

proc gplot;
plot ga*bw=sex / haxis=axis2 vaxis=axis1;
axis1 label=(a=90 'Gestational Age');
title;
footnote;
run;
* HISTOGRAM OF BIRTH WEIGHT - See bottom next page for output;

proc gchart;
vbar bw;
title 'Histogram of Birth Weight';
BIO113 71
run;
BIO113 72
ga
33
32
31
30
29
28
27
26
25
24
500 600 700 800 900 1000 1100 1200 1300 1400 1500
bw
FREQUENCY
70
60
50
40
30
20
10
0
520 600 680 760 840 920 1000 1080 1160 1240 1320 1400 1480
bw MI DPOI NT
BIO113 73
DATA MANAGEMENT I - SET: (LSB 5.1-5.3)
Basics: SET reads existing SAS data sets not 'raw' data
Reads in observations, sequentially, one at a time
May read and modify a single data set
May read, combine (concatenate) and modify multiple data sets
First, create 2 SAS data sets (work.one and work.two):
* create two SAS data sets for data management tasks;
data one; Data:

input id 1-2 ga 3-4 bw 5-8 cs 9; 1 3114621
run; 8 16500
data two; 3 25 931
infile 'p:\bio113\ex29.dat'; 6 3213270
input id ga bw cs acs;
run; 7 26 880 0 1
10 26 736 . 1
proc print data=one; title 'one'; run; 2 25 564 1 0
proc print data=two; title 'two'; run; 11 30 1212 1 0
Result:
one
Obs id ga bw cs
1 1 31 1462 1
2 8 . 1650 0
3 3 25 931 .
4 6 32 1327 0
two
Obs id ga bw cs acs
1 7 26 880 0 1
2 10 26 736 . 1
3 2 25 564 1 0
4 11 30 1212 1 0
BIO113 74
Concatenate 2 SAS data sets:
Example: Concatenate SAS data sets, work.one and work.two, with SET.
data in.baby1; set one two; * creates baby1 *;
run;
proc print; title 'one and two concatenated'; run;
Result:
one and two concatenated
Obs id ga bw cs acs
1 1 31 1462 1 .
2 8 . 1650 0 .
3 3 25 931 . .
4 6 32 1327 0 .
5 7 26 880 0 1
6 10 26 736 . 1
7 2 25 564 1 0
8 11 30 1212 1 0
Interleave 2 SAS data sets:
Example: Repeat, but this time interleave the 2 SAS data sets by ID and create the new
variable, gacat. Both data sets need to be sorted before interleaving.
proc sort data=one; by id; run;
proc sort data=two; by id; run;
data in.baby1; set one two; by id; * replaces old baby1 with new one *;
if . < ga < 30 then gacat='younger';

else if ga >= 30 then gacat='older';
run;
proc print; title 'one and two interleaved'; run;
Result:
one and two interleaved
Obs id ga bw cs acs gacat
1 1 31 1462 1 . older
2 2 25 564 1 0 younger
3 3 25 931 . . younger
4 6 32 1327 0 . older
5 7 26 880 0 1 younger
6 8 . 1650 0 .
7 10 26 736 . 1 younger
8 11 30 1212 1 0 older
BIO113 75
Create a subset:
Example: Now let's create a new temporary data set, work.older, that's a subset of the data set
in.baby1 created in the previous step. The subsetting IF statement tells SET to bring
in a subset of observations, those with gacat='older'.
data older; set in.baby1;

if gacat='older';
run;
proc print; title 'older babies';

run;
Result:
older babies
1 1 31 1462 1 . older
2 6 32 1327 0 . older
3 11 30 1212 1 0 older
Example: Now let's create a new temporary data set, work.cs, that's a different subset of
work.baby1. The subsetting IF statement tells SET to bring in the subset of
observations with non-missing cs.
data cs; set in.baby1;

if cs > .; *** could be written: if cs ^= .; ***;
run;
proc print; title 'babies not missing on cs';

run;
Result:
babies not missing on cs
1 1 31 1462 1 . older
2 2 25 564 1 0 younger
3 6 32 1327 0 . older
4 7 26 880 0 1 younger
5 8 . 1650 0 .
6 11 30 1212 1 0 older
BIO113 76
DATA MANAGEMENT II - MERGE: (LSB 5.4-5.7, 5.9-5.10, 5.14)
Basics: MERGE reads existing SAS data sets not 'raw' data
Reads in observations, sequentially, one at a time
Reads, combines (joins) and modifies multiple data sets
Meaningful joining requires one or more BY variables
Joining is one-to-one, one-to-many or many-to one but never many-to-many
First use SET to get a different subset of observations and variables from work.baby1 and then
create a second SAS data set. Notice KEEP= option and DROP statement.
* make data sets for merge examples;
data one; set in.baby1 (keep=id bw ga acs);

drop bw;
if . < bw < 1000;
run;
data two; Data:

input id momid colloid1-colloid3 acs; 2 2 9 15 16 1
run; 10 8 22 11 19 .
3 2 4 . 9 1
proc print data=one; title 'old data'; run;
proc print data=two; title 'new data'; run;
Result:
old data
Obs id ga acs
1 2 25 0
2 3 25 .
3 7 26 1
4 10 26 1
new data
Obs id momid colloid1 colloid2 colloid3 acs
1 2 2 9 15 16 1
2 10 8 22 11 19 .
3 3 2 4 . 9 1
BIO113 77
One-to-one join:
Example: Perform a one-to-one merge, matching on ID. Data sets must be sorted prior to
merge. Note RENAME statement (rename old-varname=new-varname…;).
proc sort data=one; by id; run;
proc sort data=two; by id; run;
data in.baby2; merge one two; by id;
rename colloid1=col1 colloid2=col2 colloid3=col3;
run;
proc print; title 'baby2 - merged data set'; run;
Result:
baby2 - merged data set
Obs id ga acs momid col1 col2 col3
1 2 25 1 2 9 15 16
2 3 25 1 2 4 . 9
3 7 26 1 . . . .
4 10 26 . 8 22 11 19
BIO113 78
One-to-many join:
Example: Create a SAS data set with information about mothers, sort it and in.baby2 by
MOMID, and join the two by MOMID.
data mom; Data:
input momid age; 2 38
run; 5 34
8 29
proc sort data=in.baby2; by momid; run;
proc sort data=mom; by momid; run;
proc print data=in.baby2; title 'baby2'; run;

proc print data=mom; title 'mom'; run;
data in.baby3; merge in.baby2 mom; by momid;

run;
proc print; title 'baby3 - merge baby2 and mom data'; run;
Result:
baby2
Obs id ga acs momid col1 col2 col3
1 7 26 1 . . . .
2 2 25 1 2 9 15 16
3 3 25 1 2 4 . 9
4 10 26 . 8 22 11 19
mom
Obs momid age
1 2 38
2 5 34
3 8 29
baby3 - merge baby2 and mom data
Obs id ga acs momid col1 col2 col3 age
1 7 26 1 . . . . .
2 2 25 1 2 9 15 16 38
3 3 25 1 2 4 . 9 38
4 . . . 5 . . . 34
5 10 26 . 8 22 11 19 29
BIO113 79
Join with IN= option: Creates a special temporary SAS variable that indicates whether the
data set contributed data to the current observation. In the data step, the
value of the variable is 1 if the data set contributed data to the current
observation, and 0 otherwise. The variable is available during the data
step but is not placed in the output data set.
Example: Notice that obs 1 in in.baby3 has no data from work.mom and obs 4 has no data
from in.baby2. Now repeat the merge with the IN= option.
* Merge with in= option;

data in.baby3; merge in.baby2 (in=in1) mom (in=in2); by momid;
if in1 and in2;
run;
proc print; title 'babies have moms and moms have babies'; run;
Result:
babies have moms and moms have babies
Obs id ga acs momid col1 col2 col3 age
1 2 25 1 2 9 15 16 38
2 3 25 1 2 4 . 9 38
3 10 26 . 8 22 11 19 29
BIO113 80
Look at what IN= is doing by creating variables that hold the values of the IN= variables:
* merge DOES NOT make use of IN=;

baby=in1;
mom=in2;
run;
proc print; title; run;
Obs id ga acs momid col1 col2 col3 age baby mom
1 7 26 1 . . . . . 1 0
2 2 25 1 2 9 15 16 38 1 1
3 3 25 1 2 4 . 9 38 1 1
4 . . . 5 . . . 34 0 1
5 10 26 . 8 22 11 19 29 1 1
* merge uses IN= for BOTH data sets;

if in1 and in2;
baby=in1;
mom=in2;
run;
proc print; run;
1 2 25 1 2 9 15 16 38 1 1
2 3 25 1 2 4 . 9 38 1 1
3 10 26 . 8 22 11 19 29 1 1
* merge uses IN= for baby2 data set only;

if in1;
baby=in1;
mom=in2;
run;
proc print; run;
1 7 26 1 . . . . . 1 0
2 2 25 1 2 9 15 16 38 1 1
3 3 25 1 2 4 . 9 38 1 1
4 10 26 . 8 22 11 19 29 1 1
BIO113 81
"Many-to-one join" (doesn't use MERGE!):
Example: Find how GA and BW of the CS babies differs from the average for the group. Use
SET, not MERGE, and a SAS automatic variable, _n_. Like the variable created with
IN=, _n_ is temporary and does not appear in your SAS data set.
baby1
1 1 31 1462 1 . older
2 2 25 564 1 0 younger
3 3 25 931 . . younger
4 6 32 1327 0 . older
5 7 26 880 0 1 younger
6 8 . 1650 0 .
7 10 26 736 . 1 younger
8 11 30 1212 1 0 older
Program:
data diffs; set in.baby1;

keep id ga bw;
if cs > .;
run;
proc means noprint mean;

var ga bw;
output out=avg mean=xga xbw;
run;
data diffs; set diffs;

drop _type_ _freq_;
if _n_=1 then set avg;
gadiff=xga-ga;
bwdiff=xbw-bw;
run;
proc print; title 'ga and bw differences'; run;
Result:
ga and bw differences
Obs id ga bw xga xbw gadiff bwdiff
1 1 31 1462 28.8 1182.5 -2.2 -279.5

2 2 25 564 28.8 1182.5 3.8 618.5
3 6 32 1327 28.8 1182.5 -3.2 -144.5
4 7 26 880 28.8 1182.5 2.8 302.5
5 8 . 1650 28.8 1182.5 . -467.5
6 11 30 1212 28.8 1182.5 -1.2 -29.5
BIO113 82
DATA MANAGEMENT III - OUTPUT: (LSB 5.11)
Creates one or more SAS data sets from a single SAS data set
Overrides SAS implicit output
Example: Return to baby1 and create separate SAS data sets for babies who were
delivered by C-section and those who were not.
Notice where we use the KEEP= (or DROP=) option to keep different
variables in the two datasets.
Program:
proc print data=in.baby1; title 'baby1'; run;
data cs0 (keep=id ga acs) cs1 (keep=id bw acs); set in.baby1;
if cs=0 then output cs0;

else if cs=1 then output cs1;
else delete;
run;
proc print data=cs0; title 'cs no'; run;

proc print data=cs1; title 'cs yes'; run;
Result:
baby1
1 1 31 1462 1 . older
2 2 25 564 1 0 younger
3 3 25 931 . . younger
4 6 32 1327 0 . older
5 7 26 880 0 1 younger
6 8 . 1650 0 .
7 10 26 736 . 1 younger
8 11 30 1212 1 0 olde
cs no
Obs id ga acs
1 6 32 .
2 7 26 1
3 8 . .
cs yes
Obs id bw acs
1 1 1462 .
2 2 564 0
3 11 1212 0
BIO113 83
DATA MANAGEMENT IV – CHANGE THE UNIT OF OBSERVATION (LSB 3.9-3.10, 5.12-5.14)
Make several observations from one:
Example: Each baby has 3 colloid measures. Make an observation for each non-
missing colloid value for a baby by using OUTPUT inside the DO...END
processing of an array. This changes the unit of observation from a baby
to a baby-measurement time.
data one; set in.baby3 (keep=id momid col1-col3); if momid=2;

drop momid;
run;
proc print; title 'short fat data set'; run;
data two; set one;

keep id col time;
array CC(3) col1-col3;

do time=1 to 3;
col=CC(time);
if col > . then output;
end;
run;
proc print; title 'long skinny data set'; run;
Result:
short wide data set
Obs id col1 col2 col3
1 2 9 15 16
2 3 4 . 9
long skinny data set
Obs id time col
1 2 1 9
2 2 2 15
3 2 3 16
4 3 1 4
5 3 3 9
BIO113 84
Make one observation from several:
Example: Undo what was just done! Data set must be sorted by the id variable (ID) and
the indexing variable (TIME).
* make short data set from long one;

* notice: SET with BY;
* RETAIN;
* ARRAY processed without DO...END;
* LAST.id;
proc sort data=two; by id time;
data three; set two; by id time;
drop col time;
retain col1 col2; * NOTE: col3 does NOT have to be retained *;
array DD(3) col1-col3;
DD(time)=col;
if last.id then do;

output;
do time=1 to 3;
DD(time)=.;
end;
end;
run;
proc print; title 'short wide data set again'; run;
Result:
short wide data set again
Obs id col1 col2 col3
1 2 9 15 16
2 3 4 . 9
BIO113 85
DATA ANALYSIS - PROC STEP (LSB 7.1, 7.4-7.8, SAS User's Guide: Statistics)
PROCs work on existing SAS data sets not 'raw' data

Some PROCs (e.g., print, freq, means, univariate) will PROC the entire data set (actually,
only numeric variables) unless you supply a list of variables with VAR or TABLES
statements.
Other PROCs will fail (e.g., ttest, anova, glm, plot) unless you specify variables to use
(PLOT) and/or how to use them (CLASS, MODEL).
Univariate: Examine statistics and distribution of a continuous variable
Example: Are babies receiving more fluids/kg of birth weight on day 4 than on day 1. Is
the change normally distributed? Is it significantly different from 0?
Program: * univariate example;

data one; set in.ivh1;
delta=1000*((fluid4/bw)-(fluid1/bw));
run;
proc univariate plot; id id;
var delta;
run;
Result:
The UNIVARIATE Procedure
Variable: delta
Moments
N 553 Sum Weights 553

Mean 42.9247996 Sum Observations 23737.4142
Std Deviation 30.6322999 Variance 938.337796
Skewness -0.4271125 Kurtosis 1.15482643
Uncorrected SS 1536886.21 Corrected SS 517962.463
Coeff Variation 71.362709 Std Error Mean 1.30261789
Basic Statistical Measures
Location Variability
Mean 42.92480 Std Deviation 30.63230

Median 45.39877 Variance 938.33780
Mode 44.44444 Range 229.25737
Interquartile Range 36.62313
NOTE: The mode displayed is the smallest of 2 modes with a count of 3.
Tests for Location: Mu0=0
Test -Statistic- -----p Value------
Student's t t 32.95272 Pr > |t| <.0001

Sign M 232.5 Pr >= |M| <.0001
Signed Rank S 71756.5 Pr >= |S| <.0001
BIO113 86
Quantiles (Definition 5)
Quantile Estimate
100% Max 136.26374

99% 114.98258
95% 88.60759
90% 78.78788
75% Q3 62.26415
50% Median 45.39877
25% Q1 25.64103
10% 4.90196
5% -14.28571
1% -36.78161
0% Min -92.99363
The UNIVARIATE Procedure

Variable: delta
Extreme Observations
-----------Lowest---------- ----------Highest----------
Value id Obs Value id Obs
-92.9936 110151 9 122.609 111531 110

-75.9398 136971 420 125.974 220151 493
-55.4348 111131 84 128.205 111051 77
-42.0455 110902 66 130.556 136482 389
-39.1892 136161 372 136.264 111521 109
Missing Values
-----Percent Of-----
Missing Missing
Value Count All Obs Obs
. 11 1.94 100.00
Histogram # Boxplot Normal Probability Plot

135+* 2 0 135+ *
.** 3 0 | **
.* 2 | | +*++
.*** 5 | | +***
.******* 14 | | +****
.************** 27 | | *****
.******************** 40 | | +****
.******************************* 61 +-----+ | *****
.****************************************** 84 | | | *****
.************************************** 76 *--+--* | *****
.*************************************** 77 | | | ****+
.************************** 51 +-----+ | ****
.********************* 41 | | ****
.*************** 30 | | ****
.******* 14 | | +***
.**** 8 | | ++***
.****** 11 0 | ++****
.*** 5 0 |+***
.* 1 0 |*
.* 1 0 |*
. |
.* 1 0 |*
. |
-95+* 1 * -95+*
----+----+----+----+----+----+----+----+-- +----+----+----+----+----+----+----+----+----+----+
may represent up to 2 counts -2 -1 0 +1 +2
BIO113 87
BIO113 88
T test: Compare mean value for 2 samples (independent t-test)
Example: Do babies with and without IVH have different pCO2 values on day 1 or day 2?
Program: * ttest example - use work.one from univariate example;

proc format;
value yn 0='No' 1='Yes';
run;
proc ttest;
class ivh; format ivh yn.;
var pco2_1 pco2_2;
run;
Result:
The TTEST Procedure
Statistics
Lower CL Upper CL Lower CL Upper CL

Variable Class N Mean Mean Mean Std Dev Std Dev Std Dev Std Err Minimum Maximum
pco2_1 No 425 32.289 33.179 34.069 8.7435 9.3315 10.005 0.4526 6 68

pco2_1 Yes 110 29.173 30.873 32.572 7.9408 8.9925 10.368 0.8574 15 58
pco2_1 Diff (1-2) 0.3595 2.3061 4.2527 8.7389 9.2632 9.8548 0.9909
pco2_2 No 297 35.321 36.212 37.104 7.2261 7.8076 8.4916 0.453 15 64
pco2_2 Yes 101 35.191 36.713 38.234 6.7714 7.7076 8.9465 0.7669 16 59
pco2_2 Diff (1-2) -2.263 -0.501 1.2616 7.2761 7.7824 8.3651 0.8964
T-Tests
Variable Method Variances DF t Value Pr > |t|
pco2_1 Pooled Equal 533 2.33 0.0203

pco2_1 Satterthwaite Unequal 175 2.38 0.0185
pco2_2 Pooled Equal 396 -0.56 0.5767
pco2_2 Satterthwaite Unequal 175 -0.56 0.5747
Equality of Variances
Variable Method Num DF Den DF F Value Pr > F
pco2_1 Folded F 424 109 1.08 0.6504

pco2_2 Folded F 296 100 1.03 0.8965
BIO113 89
Npar1way: Compare medians for two or more groups with the Wilcoxon rank sums test or
the Kruskal-Wallis test. Unfortunately, Npar1way does not display the
medians only the mean of the ranks for each category (mean score).
Example: Is MAP (mean blood pressure) on day 1 different for babies whose mothers
received no, partial or complete antenatal corticosteriod courses?
Program: * npar1way example - use work.one again;

proc format;
value acs 1="none" 2="partial" 3="complete";
run;
proc npar1way wilcoxon;
class acs; format acs acs.;
var map1;
run;
Result:
The NPAR1WAY Procedure
Wilcoxon Scores (Rank Sums) for Variable map1

Classified by Variable acs
Sum of Expected Std Dev Mean

acs N Scores Under H0 Under H0 Score
------------------------------------------------------------------------
partial 130 35420.50 33215.00 1448.78971 272.465385
complete 207 55222.50 52888.50 1632.48186 266.775362
none 173 39662.00 44201.50 1573.90994 229.260116
Average scores were used for ties.
Kruskal-Wallis Test
Chi-Square 8.4380
DF 2
Pr > Chi-Square 0.0147
BIO113 90
Reg: Perform univariate or multivariate linear regression. This is one of several SAS procedures
for linear regression and is nice because diagnostic plots are built in.
Example: What's the relationship between birth weight and gestational age? Is the
relationship linear?
Program: * reg example - use work.one again;

proc reg;
model bw = ga;
plot R. * ga;
run;
/* note: proc reg lineprinter; gives dot matrix plot */
Result:
The REG Procedure
Model: MODEL1
Dependent Variable: bw
Analysis of Variance
Sum of Mean
Source DF Squares Square F Value Pr > F
Model 1 17780974 17780974 535.30 <.0001

Error 560 18601510 33217
Corrected Total 561 36382484
Root MSE 182.25526 R-Square 0.4887

Dependent Mean 1016.53381 Adj R-Sq 0.4878
Coeff Var 17.92909
Parameter Estimates
Parameter Standard
Variable DF Estimate Error t Value Pr > |t|
Intercept 1 -1122.36248 92.76595 -12.10 <.0001

ga 1 77.28300 3.34031 23.14 <.0001
BIO113 91
bw = - 1122. 4 +77. 283 ga

600 N
562
Rsq
0. 4887
400 Adj Rsq
0. 4878
RMSE
182. 26
200
- 200
- 400
- 600
- 800
24 25 26 27 28 29 30 31 32 33
ga
BIO113 92
ANOVA: Perform one-way analysis of variance--compare means when there are more
than two groups. Make multiple comparisons with Bonferroni, Duncan, Tukey,
Scheffe, etc.
Example: Do the babies receive different amounts of colloid (in cc/kg birth weight) on days 1 to
4? Note we need to make a 'long skinny' data set.
Program: * 1-way anova example after making long data set;

data two; set one (keep=bw col1-col4);
array CC(4) col1-col4;
do day=1 to 4;
colloid=1000*(CC(day)/bw);
output;
end;
run;
proc anova;
class day;
model colloid=day;
means day / tukey bon;
run;
Result:
The ANOVA Procedure
Class Level Information
Class Levels Values
day 4 1 2 3 4
Number of observations 2264
NOTE: Due to missing values, only 657 observations can be used in this analysis.
The ANOVA Procedure
Dependent Variable: colloid
Sum of
Source DF Squares Mean Square F Value Pr > F
Model 3 13052.13753 4350.71251 33.59 <.0001
Error 653 84573.99588 129.51607
Corrected Total 656 97626.13341
R-Square Coeff Var Root MSE colloid Mean
0.133695 65.34327 11.38051 17.41650
Source DF Anova SS Mean Square F Value Pr > F

BIO113 93
day 3 13052.13753 4350.71251 33.59 <.0001

BIO113 94
The ANOVA Procedure
Tukey's Studentized Range (HSD) Test for colloid
NOTE: This test controls the Type I experimentwise error rate.
Alpha 0.05
Error Degrees of Freedom 653
Error Mean Square 129.5161
Critical Value of Studentized Range 3.64260
Comparisons significant at the 0.05 level are indicated by ***.
Difference
day Between Simultaneous 95%
Comparison Means Confidence Limits
1 - 2 8.431 5.527 11.335 ***

1 - 3 9.299 6.143 12.456 ***
1 - 4 10.221 6.795 13.646 ***
2 - 1 -8.431 -11.335 -5.527 ***
2 - 3 0.868 -2.492 4.229
2 - 4 1.790 -1.825 5.404
3 - 1 -9.299 -12.456 -6.143 ***
3 - 2 -0.868 -4.229 2.492
3 - 4 0.922 -2.899 4.742
4 - 1 -10.221 -13.646 -6.795 ***
4 - 2 -1.790 -5.404 1.825
4 - 3 -0.922 -4.742 2.899
The ANOVA Procedure
Bonferroni (Dunn) t Tests for colloid
NOTE: This test controls the Type I experimentwise error rate, but it generally has a
higher
Type II error rate than Tukey's for all pairwise comparisons.
Alpha 0.05
Error Degrees of Freedom 653
Error Mean Square 129.5161
Critical Value of t 2.64632
Comparisons significant at the 0.05 level are indicated by ***.
Difference
day Between Simultaneous 95%
Comparison Means Confidence Limits
1 - 2 8.431 5.448 11.415 ***

1 - 3 9.299 6.056 12.542 ***
1 - 4 10.221 6.702 13.740 ***
2 - 1 -8.431 -11.415 -5.448 ***
2 - 3 0.868 -2.585 4.321
2 - 4 1.790 -1.924 5.504
3 - 1 -9.299 -12.542 -6.056 ***
3 - 2 -0.868 -4.321 2.585
3 - 4 0.922 -3.004 4.847
4 - 1 -10.221 -13.740 -6.702 ***
4 - 2 -1.790 -5.504 1.924
BIO113 95
4 - 3 -0.922 -4.847 3.004

BIO113 96
GLM: Perform 1- to N-way analysis of variance
Example: Is the pCO2 on the first day of life associated with gestational age
(categorized), birth hospital, an interaction between the two?
Program:
* 2-way anova with interaction - use work.one & create ga categories;
data one; set one;
if . < ga < 26 then gacat=1;
else if 26 <= ga < 29 then gacat=2;
else if ga >= 29 then gacat=3;
proc glm;
class gacat hosp;
model pco2_1=gacat hosp gacat*hosp;
run;
Result:
The SAS System
The GLM Procedure
Class Level Information
Class Levels Values
gacat 3 1 2 3
hosp 2 1 2
Number of observations 566
NOTE: Due to missing values, only 533 observations can be used in this analysis.
Dependent Variable: pco2_1
Sum of
Source DF Squares Mean Square F Value Pr > F
Model 5 16328.03364 3265.60673 57.61 <.0001
Error 527 29870.72058 56.68068
Corrected Total 532 46198.75422
R-Square Coeff Var Root MSE pco2_1 Mean
0.353430 23.01959 7.528658 32.70544
Source DF Type I SS Mean Square F Value Pr > F
gacat 2 3494.42522 1747.21261 30.83 <.0001

hosp 1 12719.74043 12719.74043 224.41 <.0001
gacat*hosp 2 113.86799 56.93400 1.00 0.3669
Source DF Type III SS Mean Square F Value Pr > F
gacat 2 2574.98973 1287.49486 22.71 <.0001

hosp 1 11786.47349 11786.47349 207.95 <.0001
BIO113 97
gacat*hosp 2 113.86799 56.93400 1.00 0.3669

BIO113 98
Logist: Perform logistic regression (outcome variable is dichotomous).
Example: Is development of IVH related to characteristics of the babies and/or their

hospital experience?
Program:
* logistic regression example - use work.one and create dummy variables;
data one; set one;
if ga > . then ga1=(. < ga < 26);
if ga > . then ga2=(26 <= ga < 29);
if rom > . then romcat=(rom >= 1);
if acs > . then complete=(acs=3);
run;
proc logistic descending;
model ivh=ga1 ga2 pih romcat complete;
run;
Result:
The LOGISTIC Procedure
Model Information
Data Set WORK.ONE

Response Variable ivh
Number of Response Levels 2
Number of Observations 531
Model binary logit
Optimization Technique Fisher's scoring
Response Profile
Ordered Total
Value ivh Frequency
1 1 105
2 0 426
Probability modeled is ivh=1.
NOTE: 35 observations were deleted due to missing values for the response or
explanatory variables.
BIO113 99
Model Convergence Status
Convergence criterion (GCONV=1E-8) satisfied.
Model Fit Statistics
Intercept
Intercept and
Criterion Only Covariates
AIC 530.083 496.101

SC 534.358 521.750
-2 Log L 528.083 484.101
Testing Global Null Hypothesis: BETA=0
Test Chi-Square DF Pr > ChiSq
Likelihood Ratio 43.9819 5 <.0001

Score 42.8291 5 <.0001
Wald 38.1241 5 <.0001
Analysis of Maximum Likelihood Estimates
Standard Wald
Parameter DF Estimate Error Chi-Square Pr > ChiSq
Intercept 1 -2.0982 0.3056 47.1438 <.0001

ga1 1 1.6962 0.3398 24.9185 <.0001
ga2 1 0.9517 0.3105 9.3958 0.0022
pih 1 -0.5449 0.3898 1.9538 0.1622
romcat 1 0.2267 0.2380 0.9071 0.3409
complete 1 -0.6469 0.2461 6.9094 0.0086
Odds Ratio Estimates
Point 95% Wald

Effect Estimate Confidence Limits
ga1 5.453 2.802 10.614

ga2 2.590 1.409 4.760
pih 0.580 0.270 1.245
romcat 1.254 0.787 2.000
complete 0.524 0.323 0.848
Association of Predicted Probabilities and Observed Responses
Percent Concordant 66.4 Somers' D 0.400

Percent Discordant 26.5 Gamma 0.430
Percent Tied 7.1 Tau-a 0.127
Pairs 44730 c 0.700
BIO113 100
Lifetest: Perform Kaplan-Meier (product limit) survival analysis.
Example: Is hospital length of stay related to gestational age (categorical). Deaths are
censored
Program:
proc lifetest notable plots=(s) data=one;
time los*dead(1);
strata gacat;
run;
/* note: proc lifetest notable plots=(s) data=one lineprinter; for dot plot*/
Result:
The LIFETEST Procedure
Summary of the Number of Censored and Uncensored Values
Percent
Stratum gacat Total Failed Censored Censored
1 1 113 107 6 5.31

2 2 254 247 7 2.76
3 3 197 196 1 0.51
-------------------------------------------------------------------
Total 564 550 14 2.48
NOTE: There were 2 observations with missing values, negative time values or frequency
values less than 1.
Testing Homogeneity of Survival Curves for los over Strata
Rank Statistics
gacat Log-Rank Wilcoxon
1 -73.669 -26890
2 -19.678 -14009
3 93.346 40899
Covariance Matrix for the Log-Rank Statistics
gacat 1 2 3
1 114.047 -84.546 -29.500

2 -84.546 132.927 -48.381
3 -29.500 -48.381 77.881
Covariance Matrix for the Wilcoxon Statistics
gacat 1 2 3
1 10992531 -7436866 -3555664

2 -7436866 14619495 -7182629
3 -3555664 -7182629 10738293
Test of Equality over Strata

Pr >
Test Chi-Square DF Chi-Square
Log-Rank 126.1490 2 <.0001

BIO113 101
Wilcoxon 173.9234 2 <.0001

-2Log(LR) 62.2093 2 <.0001
BIO113 102
This graph would be in color on your screen. The line on the left is the babies in gacat=3 (29
weeks and above), the middle line is the babies in gacat=2 (26 to 28 weeks), and the line on the
right is the babies in gacat=1 (less than 26 weeks).
1. 00
0. 75
0. 50
0. 25
0. 00
0 50 100 150 200 250 30
l os
STRATA: gacat =1 Censor ed gacat =1 gacat =2

Censor ed gacat =2 gacat =3 Censor ed gacat =3

Introduction To Data Management and Programming in SAS

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Introduction To Data Management and Programming in SAS

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Introduction To Data Management and Programming in SAS

Uploaded by

Copyright:

Available Formats

BIO113 i

Introduction to Data Management and Programming in SAS

Harvard School of Public Health

General Information (LSB 1.1-1.10, 3.11, 8.1-8.2)

What is as SAS data set ...............................................................................................1

Creating SAS data sets (LSB 2.1-2.7, 2.12-2.16, 8.3-8.6, 8.13-8.14)

Begin the data step: DATA.......................................................................................... 10

External raw data sets: FILENAME and fileref ............................................................ 26

Conditional processing (LSB 3.4-3.6)

Conditional assignment: IF-THEN, IF-THEN-DO-END ............................................... 44

Arrays (LSB 3.10)

Define arrays: ARRAY ................................................................................................ 48

Sorting and printing data with PROCs (LSB 4.1-4.7)

Basics: PROC PRINT, PROC SORT .......................................................................... 55

Simple data summarization (LSB 4.9-4.11, 4.14, 7.2-7.3)

Summarize categorical data: PROC FREQ ................................................................ 61

Data management I - SET (LSB 5.1-5.3)

Data management II - MERGE (LSB 5.4-5.7, 5.9-5.10, 5.14)

Data management III - OUTPUT (LSB 5.11) ....................................................................... 80

Data management IV - Change unit of observation (LSB 3.9-3.10, 5.12-5.14)

Make several observations from one .......................................................................... 81

PROC UNIVARIATE ................................................................................................... 83

GENERAL INFORMATION (LSB 1.1-1.10, 3.11, 8.1-8.2)

Components: OPTIONS statements

OPTIONS: begins with keyword OPTIONS

Data set definition: begins with the keyword LIBNAME or FILENAME

DATA step: begins with the keyword DATA

PROC step: begins with the keyword PROC

Comments: annotate and document SAS programs

Run: execute preceding lines in DATA or PROC step

Example of a SAS program:

1 * example of a SAS program;

NOTE: The data set WORK.SASCLASS has 2 observations and 2 variables.

NOTE: There were 2 observations read from the dataset WORK.SASCLASS.

The SAS System

Obs name height

Names consist of 32 or fewer letters, numbers or underscores (_) EXCEPT dataset

SAS variables: variables are either character or numeric.

TYPE WHAT LENGTH MISSING VALUE

Lists of variables: (LSB 3.11)

Statements (except assignment statements) begin with a SAS keyword

Begin each SAS statement on a new line (but, I don’t always...)

Separate tasks in the program with blank lines

Use comments: * comment; or /* comment */

Complete each DATA and PROC step with a RUN statement

SAS Windowing Environment: Consists of five windows.

PROGRAM EDITOR: enter and modify SAS programs and data

LOG: SAS's annotation of programs after execution (black)

OUTPUT: Results of PROCedures

RESULTS: Table of contents/index of your output

EXPLORER: Access to data libraries and files

2. The SAS menu bar

File open, save and print files and exit SAS

3. The SAS tool bar (command bar and various buttons)

Command submit short commands to SAS

4. The SAS window bar

Click on a tab to see a hidden window and make it active.