Workshop 09 S1 2020 Solutions
Workshop 09 S1 2020 Solutions
Workshop 09 S1 2020 Solutions
Workshop 9 solutions
Question 9.1
It is generally agreed that convertibles offer freedom and excitement. They create a sensation
of being a part of your surroundings that no sunroof can offer. On the other hand, sedans are
stylish yet practical for small families. The back doors are also handy for rear seat access, and
the boot is a secure way to store luggage or cargo.
A car importer believes that there is a relationship between buyers' gender and their preference
of sedans or convertibles. Two importers gathered some information on the preference of male
and female buyers and this is shown in tables A and B below:
Gender
TABLE A
Male Female Total
Convertibles 30 8
Preference
Sedans 56 38
Total
Gender
TABLE B
Male Female Total
Convertibles 15 4
Preference
Sedans 28 19
Total
(a) In a test for independence between the two variables, using either one of these two tables
of observed frequencies, what is the number of degrees of freedom of the χ2 distribution
under the null hypothesis?
There are two row categories and two column categories and so we have
(2-1)(2-1) = 1 degree of freedom
1
Business Statistics
(b) Note that the sample size in Table A is twice that in Table B, and that exactly the same
proportion are in each category in each case. Conduct a test of independence for the
variables preference and gender at the 5% level of significance using Table A. Use the p-
value approach. Show all steps and workings for each step.
fo (fe) Gender
Convertibles 30 ( 38 × 86 = 24.76) 8(
38 × 46
= 13.24) 38
132 132
Sedans 94 × 86 94 × 46 94
56 ( = 61.24) 38 ( = 32.76)
132 132
Total 86 46 132
𝒇𝟎 𝒇𝒆 (𝒇𝟎 − 𝒇𝒆 )𝟐
𝒇𝒆
38 × 86
30 = 24.76 1.11
132
38 × 46
8 = 13.24 2.07
132
94 × 86
56 = 61.24 0.45
132
94 × 46
38 = 32.76 0.84
132
TOTAL 4.47
2
Business Statistics
Step 1: Hypotheses
(𝑓0 − 𝑓𝑒 )2
χ2 = ∑
𝑓𝑒
𝑎𝑙𝑙 𝑐𝑒𝑙𝑙𝑠
𝛘𝟐𝒄𝒂𝒍𝒄 = 𝟒. 𝟒𝟕
Step 3: p-value
3
Business Statistics
Step 5: Conclusion
Question 9.2
Conduct a test of independence for the variables preference and gender at the 5% level of
significance using Table B in question 9.1. Use the p-value approach. Show all steps and
workings for each step.
Construct contingency table with observed and expected frequencies
𝒇𝟎 𝒇𝒆 (𝒇𝟎 − 𝒇𝒆 )𝟐
𝒇𝒆
19 × 43
15 = 12.38 0.55
66
19 × 23
4 = 6.62 1.04
66
47 × 43
28 = 30.62 0.22
66
47 × 23
19 = 16.38 0.42
66
TOTAL 2.23
Since every cell in the contingency table has expected frequencies greater than 5, confirm χ2
distribution is appropriate.
Step 1: Hypotheses
4
Business Statistics
2
(𝑓0 − 𝑓𝑒 )2
χ = ∑
𝑓𝑒
𝑎𝑙𝑙 𝑐𝑒𝑙𝑙𝑠
𝛘𝟐𝒄𝒂𝒍𝒄 = 𝟐. 𝟐𝟑
Step 3: p-value
Step 5: Conclusion
5
Business Statistics
9.3
In question 9.1 and 9.2, you should have found that the two tests give different sample values
of χ2 , and different conclusions regarding the independence of the variables.
(a) Point out why the formula for the χ2 value allows it to be different in the two cases even
though all the observed and expected frequencies are proportional.
(f0 −fe )2
The formula for the test statistic χ2 = ∑all cells has a square of frequencies in the
fe
numerator and just a frequency in the denominator. So, if you double all the frequencies,
you will double χ2 .
(b) Explain why it is intuitively reasonable that the two tests should give different
conclusions.
If you have a very small sample, then rather big deviations from the expected frequencies
can occur without significance. It only takes a few extra in one category over another to
make a big difference in proportion. As the sample size increases, the same proportionate
difference will only occur if many data points are present above expectation, and the
6
Business Statistics
Further practice:
Q9.4
Correctly predicting the direction of change in foreign currency exchange can be lucrative. 216
investors were asked to predict the direction of change over a certain period, and the actual
direction was later recorded. The results are given in the following table.
Predicted direction:
Direction: Down 65 64
Up 39 48
TOTAL
(a) Test the hypothesis that a relationship exists between the predicted and actual rates of
exchange. Use a 10% level of significance and the p-value approach.
7
Business Statistics
Since every cell in the contingency table has expected frequencies greater than 5, confirm χ2
distribution is appropriate.
Step 1: Hypotheses
2
(𝑓0 − 𝑓𝑒 )2
χ = ∑
𝑓𝑒
𝑎𝑙𝑙 𝑐𝑒𝑙𝑙𝑠
𝛘𝟐𝒄𝒂𝒍𝒄 = 𝟎. 𝟔𝟒𝟑
Step 3: p-value
8
Business Statistics
Step 5: Conclusion
(b) Based on this data, would you say that the predictions of investors should be taken into
account in forecasting exchange rate changes?
No. Based on this data it has not been established even at the 10% level of significance
that there is a significant relationship between predictions by investors and what actually
happens.