Keywords additive mixed-effects model bias-corrected sandwich variance correlated time-to-event outcomes power analysis sample size calculation unequal cluster sizes