To increase the power of the individual tests.

To decrease the standard error of the group means.

To control the family-wise error rate by preventing an inflated chance of committing at least one Type I error.

To ensure that all group variances are equal.