Fig. 1.

The FWER (top) and disjunctive power (bottom) obtained when evaluating two continuous outcomes using a variety of methods to control the FWER. In the left hand graphs, there are no missing data. In the right hand graphs, the missing data are missing completely at random, with 15% missing in the first outcome and 25% missing in the second outcome (‘Missing data’). The graphs display various degrees of correlation between the outcomes, ranging from ρ = 0 to ρ = 0.8. The Monte Carlo standard errors (MCSE) were similar across all methods. When there were no missing data, the MCSE was between 0.002–0.004 for the disjunctive power and 0.002–0.004 for the FWER. In the missing data scenario, the MCSE was between 0.002–0.003 for the disjunctive power and between 0.003–0.005 for the FWER.)