For the data in Problem 15 in Chapter 5, use LN_BRNTL as the response variable and LN_PPMTL, LN_BLDTL, AGE, and WEIGHT as predictors. (Use α = .05.)
a. Indicate a plausible fixed order for testing predictors, based on the nature of the data. Briefly defend your choice.
b. Use a computer program to fit the full model. Provide a test of whether or not the population multiple correlation is 0.
c. Using the fixed order, choose a best model, adding variables in a forward fashion. Do not adhere to a particular a, but use your judgment.
d. Repeat part (c) but delete the variables in the fixed order (a backward method).
e. Using a computer program employing a stepwise procedure, use a backward algorithm to find the best model.
f. Repeat part (e), using a forward algorithm.
g. Compare and contrast your conclusions for parts (c), (d), (e), and (f). Indicate your preferred model.
h. Using the ideas presented in Chapter 14, indicate how you could evaluate the adequacy of the best model and the validity of the underlying assumptions.
i. Suggest a practical split-sample approach for this particular set of data. Include recommended sample sizes, any stratification variables, and a variable selection strategy.