Splus Homework Set 6, Due Monday April 28, 2003

Download the ASCII dataset

http://lib.stat.cmu.edu/datasets/Plasma_Retinol

(A) Fit a linear regression in BOTH SAS and Splus modeling RETPLASMA
as response variable, in terms of AGE, CALORIES, FAT, FIBER,
CHOLESTEROL BETADIET, RETDIET, and the factor/class variables SEX and
SMOKSTAT.

   (i) Which variables are important ? Which are found by PROC GLM or
the Splus stepwise linear-model fitting functions using a FORWARD or
BACKWARD selection  method ?

   (ii) Reproduce all of the entries in the Type I and TYPE III
Sum-of-squares table from PROC GLM using Splus functions.

   (ii) Display the residuals from the best model, as a function of
the QUETELET variable (in either SAS or Splus). Can you use the plot
alone to conclude whether the QUETELET variable belongs in the model ?

(B) Fit a logistic regression model, using PROC GENMOD and the Splus
function glm, with SEX as response variable and all other variables as
potential predictors. Find the analysis of deviance table in Splus,
and reproduce the information in it in any way you can using SAS.