9.
Estimation of a Random Variables possible value
When we have collected data from an experiment, we can now
use statistical estimation to determine the likely value of such variable
when a future observation is to be performed.
Statistical inference consists of using methods by which one
makes inferences or generalizations about a population.
Point Estimator
= a rule or formula that tells us how to calculate
an estimate based on the measurements contained
in a sample. The number that results from the
calculation is called a point estimate.
Interval Estimator
= a formula that tells us how to use sample
data to calculate an interval that estimates a
population parameter .
Estimation Formulas and applied problems
A. Estimating the mean
A.1 Case 1: x known
X Z 2
x
n
(1) The quality control manager at a light bulb factory needs to
estimate the average life of a large shipment of light bulbs. The
process standard deviation is known to be 100 hours. A random
sample of 50 light bulbs indicated a sample average life of 350
hours.
a. Set up a 90% confidence interval estimate of the true average life
of light bulbs in this shipment.
b. Set up a 95% confidence interval estimate of the true average life
of light bulbs in this shipment.
c. Tell why an observed value of 320 hours would not be unusual even
though it is outside the confidence interval you calculated.
A.2 Case 2: x not known
X t
s
n
(2)The Director of Quality of a large health maintenance
organization wants to evaluate patient waiting time at a local
facility. A random sample of 25 patients selected from the
QUANMET Estimation Theory.
page 128
appointment book. The waiting time was defined as the time
from when a patient signed in to when he or she was seen by
the doctor. The following data represent the waiting time (in
minutes)
19.5 30.5 45.6 39.8 29.6
25.4 21.8 28.6 52.0 25.4
26.1 31.1 43.1
4.9 12.7
10.7 12.1
1.9 45.9 42.5
41.3 13.8 17.4 39.0 36.6
Set up a 95 % confidence interval estimate of the population
average waiting time.
B. Estimating the difference between two means
(1-2)
( x1 x 2 ) Z
B.1 Case 1: Large independent samples
12 2 2
n1
n2
(3) It is desired to estimate the difference between the mean starting
salaries for all bachelors degrees graduates of DLSU in the
Colleges of Engineering and Computer Science during the past
year. The following information is available from the Career
Development Office:
A random sample of 40 starting salaries for Engineering
graduates produced a sample mean of P230,500 and a
standard deviation of P38,515.
A random sample of 30 starting salaries for CompSci graduates
produced a sample mean of P192,000 and a standard
deviation of P35,452.
Construct a 95% confidence interval for the difference between
mean starting salaries for graduates of the two colleges. Interpret
the interval.
B.2 Case 2: Small independent samples with equal variances
( x1 x 2 ) t S p
2
freedom
( n1 1 )S1 ( n2 1 )S 2
where S p
n1 n2 2
1
1
n1 n2
value of t/2 is based on (n1+n2-2) degrees of
QUANMET Estimation Theory.
page 129
(4) The Farm has received claims from its customers that the weight of
point-of-sale adult pigs from its two farms are different by at least 5 kgs. The resident
veterinarian decided to check the validity of this claim. He took 10 pigs from each farms
available pigs for sale, and the following data was found.
Farm
1
sample size
10
pigs
Mean weight 90.5
standard deviation
3.2
of wts
Farm
2
10
pigs
85.0
3.4
a. Is the claim valid at 5% level of significance ?
b. Estimate a 95 % confidence interval for mean difference in pig
weights between those of farm 1 and those of farm 2.
B.3 Case 3: Small independent samples with unequal variances
2
( x1 x 2 ) t
s1 s2
n1 n2
where the distribution of t/2 has degrees of
freedom
s12 s22
n1 n2
s12
n1
n1 1
integer.
s22
n2
The value of should be rounded down to nearest
n2 1
(5) Youre trying to determine if a new route from your house to
school would save you at least 10 minutes of travelling time.
You
recorded 4 weeks travelling time using the two different routes and
your data showed:
Old Route (13
times)
New Route (7
QUANMET Estimation Theory.
Mean travel
time
55.2 minutes
42.7 minutes
Std
deviation
5.2
minutes
10.3
page 130
times)
minutes
Estimate a 90 % confidence interval of the difference in travelling
times if you took the new route instead of the old one.
B.4 Case 4:
Matched pairs (Dependent samples)
Let d1, d2 .. dn represent the differences between pairwise observations
in a random sample of n matched pairs. Then the small sample
confidence interval for d=(1-2) is
Sd
where d and Sd are the mean and std.dev. of the n
2
n
sample differences.
d t
(6)
A new diet program called !Give It Up! claims to be effective in
taking out the unwanted pounds off of obese people. As a
benchmark to compare against, you used the Pritikin program.
You randomly selected 30 peoples and acquired their body weights
before and after each program. The following table shows the
data.
Construct a 95% confidence interval for the mean weight loss
under each program.
!Give It
Up!
person
1
2
3
4
5
6
7
8
9
10
11
Before
After
100
kgs
124
115
125
115
112
105
112
108
95
85
70 kgs
85
80
84
89
75
85
85
92
75
70
QUANMET Estimation Theory.
Pritikin
person
Before
After
124 kgs
70 kgs
2
3
4
5
6
7
8
9
10
11
115
125
115
85
84
75
125
115
105
112
81
68
80
84
89
75
85
92
75
70
page 131
12
13
14
15
84
75
80
96
81
68
64
75
12
13
14
15
108
95
85
96
81
68
64
75
(n 1) s 2
(n 1) s 2
2
2
2
1
C. Estimating the variance
(7) Construct a 95 % CI on the variance based on the following set of
data (the amount of Krypton gas (in milliliters) that leaked out each
time its container was dropped from 4 ft above ground):
15.5 16.8 16.7 15.4 16.4 17.5
17.8 17.5 18.3 14.5 18.1 15.7
19.5 18.6 19.7 15.8 18.2 16.8
D. Estimating the ratio of two variances
12 s12
s12
1
1
2 2
2
s2 F 1, 2 2 s2 F 1, 2
1
2
2
12 s12
s12
1
F 2 , 1
s22 F 1, 2 22 s22 2
2
(8) An investor wants to compare the risks associated with two
different computer stocks, IBM and PhilCom, where the risk of a
given stock is measured by the variation of daily prices. The
investor obtains random samples of daily price changes for IBM
and PhilCom. The sample results are summarized in the
accompanying table. Compare the risks associated with both
stocks by forming a 95 % c.i. for the ratio of the true population
variances.
IBM
n1 =21
X1 = 0.585
S1 = 0.023
E. Estimating a proportion
QUANMET Estimation Theory.
PhilCom
n2 = 21
X2 = 0.572
S2 = 0.014
p Z
pq
n
page 132
(9)
In a recent employee satisfaction survey made by
J.Rizal
Industries, 356 out of 400 employees stataed that they were very
satisfied or moderately satisfied with their jobs.
Create a 95% c.i.
estimate of the population proportion of the employees who were
satisfied with their jobs.
F.
Estimating
p1 p2 Z 2
the
difference
between
two
proportions
p1 q1 p2 q 2
n1
n2
(10) In a controversial survey of dating preferences made at the
University of the Philippines in 1992, 305 out of 500 students
surveyed claimed Good Conversationalist as the one of the most
preferred trait of a dating partner.
Furthermore, 175 out of 500
claimed sexual attractiveness as the most important trait.
Estimate the difference between the proportions of the UP
studentry who preferred good talk over good looks at a 90%
confidence level.
G. Test for Goodness of Fit
Distribution and actual data
between
hypothesized
Hypothesis: Data follows the hypothesized value.
Test statistic formula:
Where
k
o i ei 2
ei
= observed frequency of the ith interval cell (i=1 to
i 1
where
k)
oi
ei = expected frequency according to the hypothesized
distribution
= Total N x (Probability Pi for ith interval.)
= N Pi
Decision: Reject the hypothesis if X 2 > X2,k-1 value on chi-squared
distribution table with probability of error and degrees of
freedom = k-1.
Example:
QUANMET Estimation Theory.
page 133
A die is to be tested if each side occurs as equally frequent as
the others. To do this, 100 throws of the die was made and the
number of occurrences per side were recorded.
X
1
2
3
4
5
6
Frequenc
2
y
12 10 15 16 22 5
Can we say with a 5% probability of error (or 95% confidence) that the
die results follow a uniform distribution?
PRACTICE PROBLEMS:
1. The production records for an automobile manufacturer show the
following figures for production per shift.
688
656
711
677
625
703
700
701
702
688
688
667
691
694
664
630
688
547
679
703
708
688
699
697
a. Give a 95% confidence interval for the production output (cars, in
this case)
b. Give an 95% confidence interval for the standard deviation of
production output that should be known to occur.
c. What proportion of the shifts should you expect to produce 680
cars or more? Give a 95% confidence interval for this proportion.
2. A single leaf was taken from each of Luciano Tangs tobacco plants.
Each was divided in half; one half was chosen at random and
treated with preparation I and the other received preparation II.
The object of the experiment was to compare the effects of the two
preparations of mosaic virus on the number of lesions of half
leaves after a fixed period of time. For a 5% level of significance,
examine the research hypothesis that the lesions that occurred
from different preparations are significantly different.
You can do
this by making a confidence interval of the differences between
each prep.
Plant
Prep
I
1
8
1
2
0
1
9 14 38 26 15
Prep
II
6 12 32 30
QUANMET Estimation Theory.
10 11
1
0
2
25
13
18
page 134
3. DJC Video Stores wants to know how long it takes for its customers
to check-out a video rental. Suppose that you are to obtain a
random sample of 20 video checkout times (in minutes) . The
following table showed the ordered sequence of data collected:
(read the data by row)
Day 1.12 2.76 3.81 4.91 1.28 5.06 5.67 6.00
1
Day 3.79 4.54 10.2 12.4 7.16 18.1
2
8
5
2
Day 1.19 0.85 2.15 15.7 1.75 5.12
3
Treat each problem below independently. Using a 0.05 level of
significance :
a. Assuming normality, give an interval estimate of the time it
takes for a customer to check out videos.
b. Assuming normality, give an interval estimate the proportion
of customers who check out within 3 minutes.
4. Louie wore XXL sized clothes in June 2002. Today, he can
consider himself normal Large sized (Size L). He shows you a
month-by-month record of his body weight for the past year. He
always weighed himself at the beginning of each month. He wants
your opinion on certain statistical claims. He started on a diet
program September 1.
Month
Weight
June
July
Aug
Sept
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
220 216 215 210 206 200 192 187 181 172 162 155
a. Give a 90% confidence interval for the month-to-month
differences in his weight since he started on the diet program.
b. Has his average monthly weight significantly changed since
he started with the program? Compare his average weight
before the program and his average weight after the
program. Use a=0.05 Is there a significant reduction?
QUANMET Estimation Theory.
page 135
