DSC 410/510 - Multivariate Statistical Methods

Assignment 5 (deadline: 4pm on Monday, October 18)

Practical matters

Illustration from p429-436:

• Reproduce the analysis described in the book by using SAS with the "hatco2raw.xls" spreadsheet from the data page. Keep in mind the following:
  • the preferences are "Metric" (not "Metric (reflected)") since higher numbers correspond to increased utility
  • use all qualitative factors for the five attributes (even though two of them could be treated as quantitative)
  • since the last 4 of the 22 stimuli are holdout stimuli for validation (not to be used in estimation) use the "Weight" variable (move it to the "Weight" box in the variable selection dialog box) - the last 4 stimuli have zero values for "Weight" and so are not used in the analysis
  • overall relative importances of the attributes are calculated differently in SAS to in the book; however all the other results should match
• You can also use Microsoft Excel to check some results for this analysis - see "hatco2res.xls" from the data page for results for the 5 respondents from p433.
• Send an e-mail to ipardoe at lcbmail.uoregon.edu with answers to the following:
  1. which attribute did respondent 129 most care about, and which level of that attribute did they prefer?
  2. tell me the SAS market share predictions (which are a little different to those reported in the text book) under a logit model for products one (premixed, 50 apps, disinfect, biodegrade, 79c), two (concentrate, 200 apps, disinfect, biodegrade, 49c), and three (powder, 200 apps, disinfect, biodegrade, 35c):
     1. one: 5.0%, two: 25.5%, three: 69.5%
     2. one: 9.5%, two: 32.6%, three: 57.9%
     3. one: 8.6%, two: 30.9%, three: 60.5%
     4. one: 8.5%, two: 31.5%, three: 60.0%
     Hint: you need to "add rows" for products one, two, and three (click on each attribute box to select the levels), then make "inactive" the status of the experiment products (S1-S22) by clicking on "active" in each of those rows, then change the simulation model to "Logit."

Review questions

We may discuss some of the following questions in class on Tuesday October 19. I will also provide suggested solutions to allow you to check your answers.

• Page 437: questions 4 and 5.
• Question 3 provides the basis for the first project, so you might want to start thinking about this question too - project details are available now.