Skip to main content

Table 2 Evaluating the number of mixture components.

From: A mixture model approach to sample size estimation in two-sample comparative microarray experiments

π 0 pFDR power True JMB (sd) AIC (sd)
0.7 0.05 0.6 6 6 (0.3) 7 (0.5)
0.7 0.05 0.7 8 8 (0.5) 8 (0.8)
0.7 0.05 0.8 11 10 (1.1) 15 (4.0)
0.7 0.05 0.9 24 22 (2.5) 52 (22.5)
0.7 0.01 0.6 9 9 (0.7) 9 (0.5)
0.7 0.01 0.7 11 11 (0.8) 12 (1.0)
0.7 0.01 0.8 16 15 (1.8) 25 (8.4)
0.7 0.01 0.9 35 35 (3.6) 79 (34.4)
0.9 0.05 0.6 9 8 (1.1) 11 (3.8)
0.9 0.05 0.7 11 11 (2.4) 15 (5.6)
0.9 0.05 0.8 16 15 (4.1) 21 (8.5)
0.9 0.05 0.9 35 24 (7.9) 33 (13.7)
0.9 0.01 0.6 11 11 (1.9) 15 (6.3)
0.9 0.01 0.7 14 14 (3.2) 21 (8.8)
0.9 0.01 0.8 21 20 (6.3) 29 (12.7)
0.9 0.01 0.9 45 32 (11.1) 44 (21.8)
  1. True and estimated per group sample sizes for simulated data sets having π0 = 0.7 and π0 = 0.9, and for different pFDR and average power cutoffs. The reported sample size estimate is the average of 50 such estimates rounded off to the nearest integer. The standard deviation (sd) was based on the corresponding 50 data sets. For each data set the estimation method introduced in this paper was used with two different choices for g, the number of mixture components. The JMB column (from the author names) lists the result using a g as discussed in this paper. The AIC column lists the results using the AIC criterion for choosing g.
\