Fall 2005 Stat 514 Test 3 Take Home Problems
1. Data was collected on weight gain in adults over the holiday season. Two factors were studied, gender and whether or not the subject watched a lot of television ( > 4 hours/day) during the holiday. For females who watched TV the weight gain values were 3, 4, and 6 pounds. For females who did not watch TV the values were 7, 9, and 12. For males who watched TV the values were 16, 18, and 15. For males who did not watch TV the values were 10, 8, and 11.
Perform an analysis of variance using the aligned ranks method to test for the effects of gender, TV watching, and their interaction on holiday weight gain.
2. Use the
remember to change the ‘*’ values to ‘.’ for
missing values. For the model that predicts
the house price with the variables square feet, custom built, and age, calculate bootstrap
p values for the overall F test and for the square feet F test. Calculate a 95% bootstrap confidence interval for
the coefficient for square feet, using the t-pivot method with fixed-X sampling.
3. In section 8.1 of the text, it is recommended to take at least 800 bootstrap samples for variance estimation. In section 8.2, it is recommended to take 1000-5000 bootstrap samples for confidence interval construction. Why do these recommended amounts differ?