151
Comparing Cuddling Preferences
A recent poll of British and American pet owners shed light on the extent to which
pet owners like to cuddle with their pets. In many cases, pet owner s appear to prefer
cuddling with their pets more than with their partners. In this activity, you will be
exploring the following research question:
Is the percentage of British pet owners who prefer
cuddling with their dog different than the percentage of
American pet owners who prefer cuddling with their dog?
152
Bootstrap Test
To answer this question, you will carry out a bootstrap test using the data
from the files american-cuddle-20.tp3 and british-cuddle-20.tp3. Test
the statistical hypothesis that there is no difference between the
percentage of American dog owners who prefer to cuddle with their
dogs rather than their partners and the percentage for British dog
owners.
1. Carry out 500 trials of the bootstrap test assuming no differences between
the percentage of British and American dog owners who prefer to cuddle
with their dog rather than their partner. Sketch the distribution of
bootstrapped differences.
2. Compute the sample estimates for: (1) the percentage of American dog
owners who prefer to cuddle with their dog rather than their partner, (2) the
percentage of British dog owners who prefer to cuddle with their dog rather
than their partner, and (3) the difference between the two
percentages.
3. Com pute and report the p -value based on the observed result.
153
4. Based on the p -value you computed, how compatible is the observed
difference in means with the results produced by the model specified by
the statistical hypothesis? What does this suggest about the answer to the
research question? Explain.
Compatibility Intervals
Compatibility intervals can also be used to evaluate whether there are
statistical differences between two groups.
5. Use the data in the file british-cuddle-20.tp3 to provide an answer to
the research question: What percentage of American dog owners prefer
cuddling with their dog rather than with their partner after accounting for
sampling uncertainty? To do this, estimate the percentage from the
observed data, and then carry out 500 bootstrap trials to estimate the
uncertainty in the estimate. Plot the results from the 500 trials and
sketch the plot below. Make sure to label the axis.
6. Find the mean of the bootstrap distribution. Explain why you could expect
the bootstr ap distr ibution to be centered at this value by referring to the
model from your TinkerPlots T M sampler.
7. Compute the standard error (use the stdDev() function) based on this
simu lation.
154
8. Using the standard error, compute the margin of error.
9. Com pute the com patibility interval for the percentage of British dog
owners who prefer to cuddle with their dog rather than their partner.
10. Draw the range of the compatibility inter val for British dog owners who
prefer cuddling with their dog using the axis below.
155
11. Use the data from the file american-cuddle-20.tp3 to provide an answer to
the research question: What percentage of American dog owners prefer
cuddling with their dog rather than with their partner after accounting for
sampling uncertainty? To do this, estimate the percentage from the
observed data, and then carry out 500 bootstrap trials to estimate the
uncertainty in the estimate. Fill in the information from this analysis in
the table.
Sample Observed Standard Compatibility
Size Percentage Error Interval
12. Draw the range of the compatibility inter val for American dog owners on
the axis in Question #10.
Rem ember that the com patibility interval for British dog owners gives the
percentage of British dog owners who prefer cuddling with their dogs after
accounting for sampling uncertainty. Similarly the compatibility interval for
American dog owners gives the percentage of American dog owners who
prefer cuddling with their dogs after accounting for sam pling uncertainty. If
both intervals include some of the same values (the intervals overlap), it
provides evidence that the two groups could have the same level of
preference (i.e., evidence of no difference)…or at least the uncer tainty is too
great for us to differ entiate which group has the higher percentage.
13. Do the two intervals you drew in Question #10 overlap each other?
14. Explain using your drawing whether there is evidence that the percentage
of British dog owners who prefer cuddling with their dog is higher than the
percentage of Am erican dog owners who prefer cuddling with their dog.
156
Effect of Sample Size: n = 200 (100 per Group)
How would our results change if we had used a different sample size?
Bootstrap Test
Carry out a bootstrap test using the data from the files american- cuddle-
100.tp3 and british-cuddle-100.tp3.
15. Carry out 500 trials of the bootstrap test assuming no differences between
the percentage of American and British dog owners who prefer to cuddle
with their dog rather than their partner. Sketch the distribution of
bootstrapped differences.
16. Compute the sample estimates for: (1) the percentage of American dog
owners who prefer to cuddle with their dog rather than their partner, ( 2)
the percentage of British dog owners who prefer to cuddle with their dog
rather than their partner, and (3) the difference between the two
percentages.
17. Com pute and report the p -value based on the observed result.
157
18. Based on the p-value you com puted, how com patible is the observed
difference in means with the results produced by the model specified in
the null hypothesis? What does this suggest about the answer to the
research question? Explain.
19. How does the p-value for the bootstr ap test based on a sample size of 40
(20 in each group) compare to the p-value for the bootstr ap test based on a
sam ple size of 200 (100 in each group)?
Compatibility Intervals
20. Use the data from the file british-cuddle-100.tp3 to provide an answer to the
research question: What percentage of British dog owners prefer cuddling
with their dog rather than with their partner? To do this, estimate the
percentage from the observed data, and then carry out 500 bootstrap trials to
estimate the uncertainty in the estimate. Draw the range of this compatibility
compatibility interval using the axis below.
158
21. Use the data from the file american-cuddle-100.tp3 to provide an answer to
the research question: What percentage of American dog owners prefer
cuddling with their dog rather than with their partner? To do this, estimate
the percentage from the observed data, and then carry out 500 bootstrap
trials to estimate the uncertainty in the estimate. Fill in the information
from this analysis in the table.
Sample Observed Standard Compatibility
Size Percentage Error Interval
22. Draw the range of the compatibility inter val for American dog owners on
the axis in Question #20.
23. Based on whether or not the two intervals you drew in Question #20
overlap each other, explain whether there is evidence that the percentage
of British dog owners who prefer cuddling with their dog is higher than the
percentage of Am erican dog owners who prefer cuddling with their dog.
24. How is the result of the bootstrap test ( i.e., the p-value) effected by sample
size? Explain by com paring the p-values for the three bootstrap tests.
25. How does sample size affect statistical uncertainty ( i.e., the range of the
compatibility interval)?
26. Based on your answer to the previous question, how is statistical uncertainty
related to p-value?
