In clinical trials with repeated measurements, multiple comparisons are required to test the treatment difference at multiple selected time points. It is natural to treat the vector of test statistics as a multivariate normal or multivariate t random variable. However, numerical computation of power and sample sizes for this problem is often difficult. Specifically, when the correlations between the responses are considered, the traditional sample size calculation formulation will overestimate the sample size. So, the calculation of the effective sample size which can reach the specific power becomes critical. In this project, we developed a methodology to calculate the power function under the situation of correlated responses. We use our methods to iteratively calculate the effective sample size using multivariate normal and multivariate t distributions.