STAT460 - Experimental Design
Daily Homework #5 - Reference Distributions Review

Computational exercise:

This assignment uses the KATRINA.XLS data set. In class we analyzed the stock price data. Now, you are asked to conduct similar analyses on the share volume data. As was discussed in class, we will use a sample size of n=5 (corresponding to the number of days in the market’s week), and will only use the data for calendar year 2005 (to a void using potentially non-comparable data). In each of the following, you are answering the research question: Did Hurricane Katrina affect the volume of blue-chip stock shares traded?

1) What difference in average share volume was observed, between the five days BEFORE Katrina and the five days AFTER Katrina?

2) How likely is it to observe that sort of difference, simply by chance alone? Begin by analyzing the data using BHH’s first procedure – that is, external reference distribution with no assumption of independence. What is your p-value?

3) Ditto, only now via BHH’s second procedure: nonoverlapping differences of means, to give (essentially) independent samples.

4) Ditto, via BHH’s third procedure.

5) Ditto, via BHH’s fourth procedure.

Review questions:

1) What is a reference distribution? Why is it important?

2) The data analysis on the Hurricane Katrina data follows the same basic patterns as the analysis of BHH’s data. However, there are some key differences in the underlying experimental situation. What are these differences and why do they matter?

3) If we use an external reference distribution in our analysis, what criteria govern our selection of which observations are to be included in that distribution?

4) What are the relative advantages and disadvantages of the use of external rather than internal reference distributions?