(Taken from Rao 16.8, p. 765) An experiment evaluates the taste of three brands of ice cream under

dierent storage conditions. The expt. calls for 12 coolers to be randomized to the 3 brands (4 coolers

per brand). Within each cooler, there is room for three half-gallon cartons of ice cream, all from the

same brand. One carton was sampled at random at the end of one day, then at the end of one week

and then one month. Each carton was evaluated for

avor by a panel of tasters. These panel scores

can be found in the le entitled \icecreamstorage.dat".

(a) This is a split-plot experiment. What are the whole plot units?

(b) What is/are the whole plot factor(s)?

(c) What is/are the sub plot factor(s)?

(d) Test for an interaction between storage time and brand. (Use = 0:05.)

(e) Test for main eects of storage time. Carry out pairwise comparisons among the three storage

time tastes, using a Bonferroni correction to control the exptwise error rate at .05.

(f) Test for main eects of brand. Report the standard error on any pairwise dierence among brands.

(g) Suppose you had averaged the data over storage time, reducing the total number of observations

from 36 down to 12. Would the F-ratio for brand eect in a a one-way ANOVA be the same as

the F-ratio from part (g)? Try it and see.