Figure 5.—

Power estimation and comparison of the EM-adjusted t-test and the ordinary t-test when P(A → a) = 10%. The points plotted indicate the empirical proportion of tests (by use of a nominal level α = 0.05) that rejected the H₀ among 1000 simulated data sets. K = Δ/(σ/n). Plots on the left correspond to cases with 10% missing data. Plots on the right correspond to cases with 20% missing data. * indicates those cases where P < 0.05, and ** indicates those cases where P < 0.005. Here “P” refers to the P-value of Wilcoxon rank-sum tests comparing the power difference between the EM-adjusted t-test and the ordinary t-test. Solid diamonds denote the power of the EM-adjusted t-test; solid squares denote the power of the ordinary t-test.