The problem considered here is to compare several methods for ANOVA under different conditions.

**Method 1**:

Perform Levene's test for homogeneity of variance. If this test does not reject the null
hypothesis, then perform ANOVA. If the F-test does not reject then say all means are equal. If
the F-test rejects, then perform pairwise t-tests with no adjustment of p-values. If Levene's test
rejects the null hypothesis, then perform pairwise t-tests with adjusted p-values method
*holm*.

**Method 2**:

Same as Method 1 except use p-value adjustment method *fdr* instead of *holm*

**Method 3**: perform pairwise t-tests with adjusted p-values method *holm*.

**Method 4**: perform pairwise t-tests with adjusted p-values method *fdr*.

**Method 5**:

Perform Kruskal-Wallis test (*kruskal.test*). If this tests does not reject, then say all
means are equal. If the Kruskal-Wallis test rejects, then perform pairwise rank sum tests
(*pairwise.wilcox.test*) with no adjustment of p-values.

Link to scripts and data is here:

http://www.utdallas.edu/~ammann/stat6341proj

2017-12-10