STA 301F97 Assignment 11: Quiz in tutorial Friday Nov 28

This assignment is based on the pulse data described in Assignment Five. For the quiz, bring your unedited list file, with NOTHING written on it except your name and student number.

1. Do a two-way analysis of variance in which the dependent variable is rise in pulse rate, and the independent variables are whether the person ran in place or not, and usual level of physical activity. Do means with the crossbreak option to see what's happening. Based on the information you have so far, describe each significant effect (except the pooled test for main effects), using plain plain language. Just say what appears to be happening, based on the means. Also, consider these questions.

• Is it fair to say that the effect of running in place appears to depend on the person's usual level of physical activity?
• Do you think it makes sense to discuss the main effect for usual level of physical activity? Why or why not?
• Does it make sense to discuss the main effect for running in place? Why or why not?
• Would you be willing to conclude that running in place caused variation in pulse rate? Why or why not?
• Would you be willing to conclude that usual level of physical activity caused variation in pulse rate? Why or why not?

2. Use the word "cell" to describe a combination of independent variable values; there are 6 cells in the two-way design we are analyzing. Use Tukey tests (from ONEWAY) to decide which cell means are different from each other. Describe what happened in plain language. Does this support your interpretation of the interaction? Does this make you change any of your answers to the other questions above? (It did for me.)

3. Look at the F-test from ONEWAY. Can you find the same test in the two-way output?

4. Now do a planned comparison that logically should be the same as the main effect for running in place. We'd expect F = t-squared (when equal variances are being assumed), but notice that it didn't happen. This is because of SPSS's screwy "classical experimental" default approach to testing for main effects. Do your ANOVA command again, this time with a / method = unique subcommand, so that all effects are adjusted for all other effects. Is that better? (Use a calculator.)