Stat 411/511

Quiz 4 Study Guide

Reading: Sleuth 5 & 6 (Optional OpenIntro 5.5)

Practice Questions: Conceptual Exercises in Sleuth
Page 141: 1, 4, 5, 6, 7, 8, 9, 11 and 12
Page 170: 6, 7, 8, 9, 10 & 11

State the null and alternative hypothesis in the one-way ANOVA.

What is a residual?

Know the steps to fill in a partially completed ANOVA table (there are many ways it could be partially completed, practice with ex 17 pg 142 in Sleuth or 5.39(c) pg 259 in OpenIntro).

How can you find the pooled standard deviation from the ANOVA table?

What are the degrees of freedom associated with the total, between groups and within groups sums of squares in a one-way ANOVA?

List the assumptions of the one-way ANOVA F-test. Which is it robust to and under what conditions?

Construct a 95% CI or do a t-test, for the difference in means for two groups in a multiple group analysis.

Name a resistant version of the one-way ANOVA. What is the null hypothesis?

Name a test to examine the equal population standard deviations assumption. What is the null hypothesis?

Why do we calculate the pooled standard deviation over all the groups even when we just want to compare two? Are there situations where this isn’t appropriate?

Can hypotheses other than the equal means model be tested using the Extra Sum of Squares F-test? Give an example.

Describe two residual plots and which assumption they examine.

Be able to set up linear comparisons on means and averages of means. Practice with ex 18, page 171 in Sleuth.

Construct an estimate and 95% confidence interval for a linear combination of means given a table of summary statistics.

Test whether a linear combination of means is equal to zero.

Define the individual error rate, and the familywise error rate.

What is the problem with looking at a lot of tests simultaneously?

Will the Tukey-Kramer confidence intervals be wider or narrower than the least significant difference confidence intervals?

What is the Bonferroni adjusted individual significance level to limit the familywise error rate at 5%?

What is wrong with examining your data to find the largest difference in two group averages and only presenting an unadjusted test for those two groups?

Because of the equivalence between confidence intervals and p-values, any adjusted confidence interval corresponds to an adjusted p-value. Will the adjusted p-value based on a Tukey-Kramer confidence interval be smaller or larger than an unadjusted p-value?