Lecture 2: More Data Structures
Statistical Computing, 36-350
Wednesday September 2, 2015
Outline
• Arrays
• Matrices
• Lists
• Data frames
• Structures of structures
Vector structures, starting with arrays
Many data structures in R are made by adding bells and whistles to vectors, i.e., they are “vector structures”
Most useful: arrays
x = c(7, 8, 10, 45)
x.arr = array(x, dim=c(2,2))
x.arr
## [,1] [,2]
## [1,] 7 10
## [2,] 8 45
dim says how many rows and columns; filled by columns
Can have 3, 4, . . . arrays; dim is vector of arbitrary length
Some properties of our array:
dim(x.arr)
## [1] 2 2
is.vector(x.arr)
## [1] FALSE
is.array(x.arr)
## [1] TRUE
1
typeof(x.arr)
## [1] "double"
str(x.arr)
## num [1:2, 1:2] 7 8 10 45
attributes(x.arr)
## $dim
## [1] 2 2
typeof() returns the type of the array elements
str() gives the structure: here, a numeric array, with two dimensions, both indexed 1–2, and then the actual
numbers
Exercise: try all these with x
Accessing and indexing arrays
Can access a 2d array either by pairs of indices or by the underlying vector:
x.arr[1,2]
## [1] 10
x.arr[3]
## [1] 10
Omitting an index means “all of it”:
x.arr[c(1:2),2]
## [1] 10 45
x.arr[,2]
## [1] 10 45
Functions on arrays
Many functions applied to a vector-structure like an array will just boil things down to the underlying vector:
2
which(x.arr > 9)
## [1] 3 4
This happens unless the function is set up to handle arrays specifically
Many functions do preserve array structure:
y = -x
y.arr = array(y,dim=c(2,2))
y.arr + x.arr
## [,1] [,2]
## [1,] 0 0
## [2,] 0 0
Others specifically act on each row or column of the array separately:
rowSums(x.arr)
## [1] 17 53
(We will see a lot more of this idea soon)
Example: houses prices in Pennsylvania
Census data for California and Pennsylvania on housing prices, by Census “tract”
calif_penn = read.csv("http://www.stat.cmu.edu/~cshalizi/uADA/13/hw/01/calif_penn_2011.csv")
penn = calif_penn[calif_penn[,"STATEFP"]==42,]
coefficients(lm(Median_house_value ~ Median_household_income, data=penn))
## (Intercept) Median_household_income
## -26206.564325 3.651256
Fit a simple linear model, predicting median house price from median household income
It turns out census tracts 24–425 are Allegheny county
Tract 24 has a median income of $14,719; actual median house value is $34,100; is that above or below what’s
predicted?
3
34100 < -26206.564 + 3.651*14719
## [1] FALSE
Tract 25 has income $48,102 and house price $155,900
155900 < -26206.564 + 3.651*48102
## [1] FALSE
What about tract 26?
We could just keep plugging in numbers like this, but that’s
• boring and repetitive
• error-prone (what if I forget to change the median income, or drop a minus sign from the intercept?)
• obscure if we come back to our work later (what are these numbers, again?)
Use variables and names
penn.coefs = coefficients(lm(Median_house_value ~ Median_household_income, data=penn))
penn.coefs
## (Intercept) Median_household_income
## -26206.564325 3.651256
allegheny.rows = 24:425
allegheny.medinc = penn[allegheny.rows,"Median_household_income"]
allegheny.values = penn[allegheny.rows,"Median_house_value"]
allegheny.fitted = penn.coefs["(Intercept)"] +
penn.coefs["Median_household_income"]*allegheny.medinc
plot(x=allegheny.fitted, y=allegheny.values,
xlab="Model-predicted median house values",
ylab="True median house values",
xlim=c(0,5e5), ylim=c(0,5e5))
abline(a=0, b=1, col="red")
4
4e+05
True median house values
2e+05
0e+00
0e+00 1e+05 2e+05 3e+05 4e+05 5e+05
Model−predicted median house values
Running example: resource allocation
Factory makes cars and trucks, using labor and steel
• a car takes 40 hours of labor and 1 ton of steel
• a truck takes 60 hours and 3 tons of steel
• resources: 1600 hours of labor and 70 tons of steel each week
Matrices
In R, a matrix is a specialization of a 2d array
factory = matrix(c(40,1,60,3), nrow=2)
factory
## [,1] [,2]
## [1,] 40 60
## [2,] 1 3
is.array(factory)
## [1] TRUE
is.matrix(factory)
## [1] TRUE
5
could also specify ncol; to fill by rows, use byrow=TRUE
Elementwise operations with the usual arithmetic and comparison operators (e.g., factory/3)
Compare whole matrices with identical() or all.equal()
Matrix multiplication
Has its own special operator, written %*%:
six.sevens = matrix(rep(7,6), ncol=3)
six.sevens
## [,1] [,2] [,3]
## [1,] 7 7 7
## [2,] 7 7 7
factory %*% six.sevens # [2x2] * [2x3]
## [,1] [,2] [,3]
## [1,] 700 700 700
## [2,] 28 28 28
(What happens if you try six.sevens %*% factory?)
Multiplying matrices and vectors
Numeric vectors can act like proper vectors:
output = c(10,20)
factory %*% output
## [,1]
## [1,] 1600
## [2,] 70
output %*% factory
## [,1] [,2]
## [1,] 420 660
(R silently casts the vector as either a 1-column or 1-row matrix, as appropriate)
Matrix operators
Transpose:
6
t(factory)
## [,1] [,2]
## [1,] 40 1
## [2,] 60 3
Determinant:
det(factory)
## [1] 60
The matrix diagonal
The diag() function can be used to extract the diagonal entries of a matrix:
diag(factory)
## [1] 40 3
It can also be used to change the diagonal:
diag(factory) = c(35,4)
factory
## [,1] [,2]
## [1,] 35 60
## [2,] 1 4
Re-set it for later:
diag(factory) = c(40,3)
Creating a diagonal or identity matrix
diag(c(3,4))
## [,1] [,2]
## [1,] 3 0
## [2,] 0 4
diag(2)
## [,1] [,2]
## [1,] 1 0
## [2,] 0 1
(How do you get a 1 x 1 matrix containing a single entry 2?)
7
Inverting a matrix
solve(factory)
## [,1] [,2]
## [1,] 0.05000000 -1.0000000
## [2,] -0.01666667 0.6666667
factory %*% solve(factory)
## [,1] [,2]
## [1,] 1 0
## [2,] 0 1
Why is it called “solve” anyway?
Solving the linear system Ax = b for x:
available = c(1600,70)
solve(factory,available)
## [1] 10 20
factory %*% solve(factory,available)
## [,1]
## [1,] 1600
## [2,] 70
Names in matrices
We can name either rows or columns or both, with rownames() and colnames()
These are just character vectors, and we use the same function to get and to set their values
Names help us understand what we’re working with
Names can be used to coordinate different objects
rownames(factory) = c("labor","steel")
colnames(factory) = c("cars","trucks")
factory
## cars trucks
## labor 40 60
## steel 1 3
8
available = c(1600,70)
names(available) = c("labor","steel")
output = c(20,10)
names(output) = c("trucks","cars")
factory %*% output # But we've got cars and trucks mixed up!
## [,1]
## labor 1400
## steel 50
factory %*% output[colnames(factory)]
## [,1]
## labor 1600
## steel 70
all(factory %*% output[colnames(factory)] <= available[rownames(factory)])
## [1] TRUE
Note that last lines don’t have to change if we add motorcycles as output or rubber and glass as inputs
(abstraction again)
Doing the same thing to each row or column
Take the mean: rowMeans(), colMeans(), input is matrix, output is vector. Also rowSums(), colSums
summary(): vector-style summary of column
colMeans(factory)
## cars trucks
## 20.5 31.5
summary(factory)
## cars trucks
## Min. : 1.00 Min. : 3.00
## 1st Qu.:10.75 1st Qu.:17.25
## Median :20.50 Median :31.50
## Mean :20.50 Mean :31.50
## 3rd Qu.:30.25 3rd Qu.:45.75
## Max. :40.00 Max. :60.00
9
apply(), takes 3 arguments:
• the array or matrix,
• then 1 for rows and 2 for columns,
• then a name of the function to apply to each
rowMeans(factory)
## labor steel
## 50 2
apply(factory, 1, mean)
## labor steel
## 50 2
(What would apply(factory, 1, sd) do?)
Lists
Sequence of values, not necessarily all of the same type
my.distribution = list("exponential", 7, FALSE)
my.distribution
## [[1]]
## [1] "exponential"
##
## [[2]]
## [1] 7
##
## [[3]]
## [1] FALSE
Most of what you can do with vectors you can also do with lists
Accessing pieces of lists
Can use [ ] as with vectors
Or use [[ ]], but only with a single index
[[ ]] drops names and structures, [ ] does not
my.distribution[2]
## [[1]]
## [1] 7
10
my.distribution[[2]]
## [1] 7
my.distribution[[2]]^2
## [1] 49
(What happens if you try my.distribution[2]ˆ2?) (What happens if you try [[ ]] on a vector?)
Expanding and contracting lists
Add to lists with c() (also works with vectors):
my.distribution = c(my.distribution,7)
my.distribution
## [[1]]
## [1] "exponential"
##
## [[2]]
## [1] 7
##
## [[3]]
## [1] FALSE
##
## [[4]]
## [1] 7
Chop off the end of a list by setting the length to something smaller (also works with vectors):
length(my.distribution)
## [1] 4
length(my.distribution) = 3
my.distribution
## [[1]]
## [1] "exponential"
##
## [[2]]
## [1] 7
##
## [[3]]
## [1] FALSE
11
Pluck out all but one piece of a list (also works with vectors):
my.distribution[-2]
## [[1]]
## [1] "exponential"
##
## [[2]]
## [1] FALSE
(What happens if you try my.distribution[[-2]]?)
Naming list elements
We can name some or all of the elements of a list:
names(my.distribution) = c("family","mean","is.symmetric")
my.distribution
## $family
## [1] "exponential"
##
## $mean
## [1] 7
##
## $is.symmetric
## [1] FALSE
my.distribution[["family"]]
## [1] "exponential"
my.distribution["family"]
## $family
## [1] "exponential"
Lists have a special shortcut way of using names, with $:
my.distribution[["family"]]
## [1] "exponential"
12
my.distribution$family
## [1] "exponential"
Names in lists (continued)
Creating a list with names:
another.distribution = list(family="gaussian",
mean=7, sd=1, is.symmetric=TRUE)
Adding named elements:
my.distribution$was.estimated = FALSE
my.distribution[["last.updated"]] = "2015-09-01"
Removing a named list element, by assigning it the value NULL:
my.distribution$was.estimated = NULL
Key-value pairs
Lists give us a natural way to store and look up data by name, rather than by position
A really useful programming concept with many names: key-value pairs, dictionaries, associative arrays
If all our distributions have components named family, we can look that up by name, without caring where
it is (in what position it lies) in the list
Data frames
The classic data table, n rows for cases, p columns for variables
Lots of the really-statistical parts of R presume data frames
Not just a matrix because columns can have different types
Many matrix functions also work for data frames (e.g.,rowSums(), summary(), apply())
(But no matrix multiplication with data frames, even if all columns are numeric!)
a.matrix = matrix(c(35,8,10,4), nrow=2)
colnames(a.matrix) = c("v1","v2")
a.matrix
## v1 v2
## [1,] 35 10
## [2,] 8 4
13
a.matrix[,"v1"] # Try a.matrix$v1 and see what happens
## [1] 35 8
a.data.frame = data.frame(a.matrix,logicals=c(TRUE,FALSE))
a.data.frame
## v1 v2 logicals
## 1 35 10 TRUE
## 2 8 4 FALSE
a.data.frame$v1
## [1] 35 8
a.data.frame[,"v1"]
## [1] 35 8
a.data.frame[1,]
## v1 v2 logicals
## 1 35 10 TRUE
colMeans(a.data.frame)
## v1 v2 logicals
## 21.5 7.0 0.5
Adding rows and columns
We can add rows or columns to an array or data frame with rbind() and cbind(), but be careful about
forced type conversions
rbind(a.data.frame,list(v1=-3,v2=-5,logicals=TRUE))
## v1 v2 logicals
## 1 35 10 TRUE
## 2 8 4 FALSE
## 3 -3 -5 TRUE
rbind(a.data.frame,c(3,4,6))
## v1 v2 logicals
## 1 35 10 1
## 2 8 4 0
## 3 3 4 6
14
Structures of structures
So far, every list element has been a single data value
List elements can be other data structures, e.g., vectors and matrices:
plan = list(factory=factory, available=available, output=output)
plan$output
## trucks cars
## 20 10
Internally, a data frame is basically a list of vectors (all of the same length)
List elements can even be other lists
which may contain other data structures
including other lists
which may contain other data structures . . .
This recursion lets us build arbitrarily complicated data structures from the basic ones
Most complicated objects are (usually) lists of data structures
Example: eigen-decomposition
eigen() finds eigenvalues and eigenvectors of a matrix
Returns a list of a vector (the eigenvalues) and a matrix (the eigenvectors)
eigen(factory)
## $values
## [1] 41.556171 1.443829
##
## $vectors
## [,1] [,2]
## [1,] 0.99966383 -0.8412758
## [2,] 0.02592747 0.5406062
class(eigen(factory))
## [1] "list"
With complicated objects, you can access parts of parts (of parts . . . )
15
factory %*% eigen(factory)$vectors[,2]
## [,1]
## labor -1.2146583
## steel 0.7805429
eigen(factory)$values[2] * eigen(factory)$vectors[,2]
## [1] -1.2146583 0.7805429
eigen(factory)$values[2]
## [1] 1.443829
eigen(factory)[[1]][[2]] # NOT [[1,2]]
## [1] 1.443829
Summary
• Arrays add multi-dimensional structure to vectors
• Matrices act like you’d hope they would
• Lists let us combine different types of data
• Data frames are hybrids of matrices and lists, allowing each column to have a different basic type
• Recursion lets us build complicated data structures out of simpler ones
16
