Skip to main content
. 2014 Sep 23;15(9):465. doi: 10.1186/s13059-014-0465-4

Table 2.

Comparing the type I error rate for the four different methods with 200 outliers

Nominal P value
Test method 0.001 0.01 0.05 0.1
Bartlett 0.01379 0.02601 0.06781 0.1179
F test 0.01467 0.02663 0.06767 0.1172
DiffVar (abs) 0.0009701 0.01006 0.05085 0.1019
DiffVar (sq) 0.0005373 0.008198 0.04880 0.1011

Median type I error rates are reported for 1000 simulations with no differentially variable or differentially methylated features, but with 200 outliers incorporated in the data. The standard deviation with which the error rate is estimated ranges from approximately 0.0003 for rates near 0.001 to 0.003 for rates near 0.1 with the Bartlett test and F test more variable than DiffVar at nominal P values of 0.001 (standard deviation approximately 0.002 compared to 0.0003) and 0.01 (standard deviation approximately 0.0018 compared to 0.0010). abs, absolute deviations; sq, squared deviations.