Question: In an experiment, rats were fed a stock ration for 100 days with various amounts of gossypol added. The following weight gains in grams were recorded.
Suppose we assume the normal regression model for these data and treat gossypol as a categorical predictor taking five levels.
(a) Create a side-by-side boxplot graph for the data. Does this give you any reason to be concerned about the assumptions that underlie an analysis based on the normal regression model?
(b) Produce a plot of the standardized residuals against the factor gossypol (if you are using a computer for your calculations, also produce a normal probability plot of the standardized residuals). What are your conclusions?
(c) Carry out a one-way ANOVA to test for any differences among the mean responses for the different amounts of gossypol.
(d) Compute 0.95-confidence intervals for all the pairwise differences of means and summarize your conclusions.