To increase the power of each individual test.

To decrease the standard deviation of the sample means.

To control the family-wise error rate by adjusting the significance level $\alpha$ for each test.

To ensure the sample size is large enough for a normal approximation.