The data listed in the following table relate to a study by Reiter and others (1981) concerning the effects of injecting triethyl-tin (TET) into rats once at age 5 days. The animals were injected with 0, 3, or 6 mg per kilogram of body weight. The response was the log of the activity count for 1 hour, recorded at 21 days of age. The rat was left to move about freely in a figure 8 maze. Analysis of other studies with this type of activity count confirms that log counts should yield Gaussian errors if the model is correct.
a. Tabulate the DOSAGE X SEX cell sample sizes. Explain why this might be called a nearly orthogonal design.
b. Using the accompanying SAS computer output, conduct a two-way ANOVA with SEX and DOSAGE as factors. (First provide the F statistic values and P-values in the accompanying ANOVA table.)
c. Using a = .05, report your conclusions based on the ANOVA.
d. Which, if any, families of means should be followed up with multiple-comparison tests? What type of comparisons would you recommend?