QMT 109: Business Research

and Statistical Methods

Wilson Gan
Lecture Ten
Reference for Definitions: Bowerman et al. (2009)
Chi-Square Goodness of Fit Test
Multinomial Experiments
Wilson Gan 2010
What are multinomial experiments?
1. Carry out n identical trials with k possible outcomes of
each trial
2. Probabilities are denoted p
1
, p
2
, , p
k
where p
1
+ p
2
+
+ p
k
= 1
3. The trials are independent
4. The results are observed frequencies of the number of
trials that result in each of k possible outcomes,
denoted f
1
, f
2
, , f
k

H
0
: multinomial probabilities are p
1
, p
2
, , p
k
H
a
: at least one of the probabilities differs from p
1
, p
2
, ,
p
k
STATEMENTS OF HYPOTHESES
Wilson Gan 2010
Chi-Square Goodness of Fit Test

k
i
i
i i
E
E f
=
1
2
2
) (

FORMULA FOR THE TEST STATISTIC

FORMULA FOR DEGREE OF FREEDOM
REJECTION RULE
2
>

2

p 1 where
p = number of treatments
f
i
= frequency of the i treatment
E
i
= np
i
= expected number of the i treatment if p
i
is
probability of being in treatment i

Wilson Gan 2010
Exercise: Wheat Bread
A large supermarket conducted a consumer preference
study by recording the brand of wheat bread purchased
by customers in its stores. The supermarket carries four
brands of wheat bread. In a random sample of 200
purchasers, the number of purchasers preferring Brands
A, B, C and D of the wheat bread were, respectively, 51,
82, 27 and 40.
Test the null hypothesis that the four brands are equally
preferred by setting = 0.05.
Find a 95 percent confidence interval for the proportion
of all purchasers who prefer Brand B.
Source: Bowerman (2009)
Wilson Gan 2010
Exercise: US Autos
The shares of the US automobile market held in 1990 by
General Motors, Japanese manufacturers, Ford, Chrysler
and other manufacturers were, respectively, 36%, 26%,
21%, 9%, and 8%. Suppose that a new survey of 1,000
new-car buyers shows the following purchase
frequencies:




Test to determine whether the current market shares
differ from those of 1990. Use = 0.05.
Source: Bowerman (2009)
GM Japanese Ford Chrysler Other
193 384 170 90 163
Chi-Square Test of Independence
The Chi Square table will be your second to the last new
table for the semester!
Wilson Gan 2010
A Motivating Example: Wearing Ateneo
Clothing
A recent graduate is interested in applying what he learned in
QMT 11. He intends to prove the obvious that Ateneo alumni
are more inclined to wear Ateneo clothing compared to those
who did not graduate from the Ateneo. He limited his sample
to 100 people from the offices of multinational firms along the
Makati Central Business District, and asked them the following
questions:
Do you have 1 or more pieces of Ateneo clothing?
Are you an Ateneo alumnus / alumnae?
Results as follows:
Owns Ateneo
Clothing?
Atenean Not an Atenean
Yes 17 7
No 8 68
Wilson Gan 2010
Another Example: The Eye Glass Effect
Do people who wear eyeglasses appear more intelligent?
25 subjects were each shown the same 4 pictures of
average-looking people two with eye glasses and two
without. The subjects were then asked to classify whether the
person in the picture seemed to have above average, average
or below average intelligence. Here is a summary of the
results:
Above
Average
Average Below
Average
With Eye
Glasses
28 20 2
Without Eye
Glasses
18 22 10
Wilson Gan 2010
Chi Square Test of Independence: Testing
independence of classifications
H
0
: the two classifications are statistically independent
H
a
: the two classifications are statistically dependent

Steps to test independence
1. Each of n randomly selected items is classified on two
dimensions into a contingency table with r rows an c columns
and let
f
ij
= observed cell frequency for i
th
row and j
th
column
r
i
= i
th
row total
c
j
= j
th
column total
2. Calculate expected cell frequency for i
th
row and j
th
column under
independence
3. Calculate the chi-square test statistic (next slide)

n
c r
E
j i
ij

Source: Bowerman (2009)

Wilson Gan 2010
The Chi-Square Test Statistic
Reject H
0
if
2
>

2
or if p-value <

2
and the p-value are based on (r-1)(c-1) degrees of
freedom

cells all
2
2

(
ij
ij ij
E
E f
=
Source: Bowerman (2009)
Ensure that sample size is large enough to safeguard test
validity
n is large enough if all expected cell frequencies are larger
than five

Wilson Gan 2010
Exercise: Wine and Tennis
A market research firm wishes to study the relationship
between wine consumption and whether a person likes to
watch professional tennis on television. One hundred
randomly selected people are asked whether they drink wine
and whether they watch tennis.
a) Test the hypothesis that whether people drink wine is
independent of whether people watch tennis. Set = 0.05.
b) Given the results of the chi-square test, does it make
sense to advertise wine during a televised tennis match
(assuming the ratings for the tennis match are high
enough)? Explain.
Watch Tennis Do Not Watch Totals
Drink Wine 16 24 40
Do Not Drink Wine 4 56 60
Totals 20 80 100
Source: Bowerman (2009)
Wilson Gan 2010
Exercise: TV Viewership Study
A television station wishes to study the relationship between
viewership of its 11:00 pm news and viewer age. A sample of
250 television viewers in each age group is randomly selected
and the number who watch the stations 11 pm news is
gathered.
Test for independence at = 0.05. Let p1, p2, p3 and p4 be
the proportions of all viewers in each age group who watch
the television stations 11:00 pm news.
Watch the 11
pm news?
18 or less 19 to 35 36 to 54 55 or older Total
Yes 37 48 56 73 214
No 213 202 194 177 786
Total 250 250 250 250 1,000
Source: Bowerman (2009)