Gravar-mail: Effects of correlation and missing data on sample size estimation in longitudinal clinical trials