Table 3.
Mean proportion false discoveries over 50 datasets for LRT and Wald statistics from the Negative Binomial mixed model when using the asymptotic 
(“LRT” and “W”, respectively) and their corresponding bootstrap-based null distribution (“LRT-boot” and “W-boot”, respectively) for testing 
versus 
or 
. We have made comparisons with the 
-test statistic in limma-voom approach (“F-voom”) and LRT and 
-test statistic in edgeR (“LRT-edgeR” and “W-edgeR”, respectively)
|
LRT | LRT-edgeR | F-edgeR | W | LRT-boot | W-boot | F-voom |
|---|---|---|---|---|---|---|---|
| 0.5 | 0.017 | 0.000 | 0.000 | 0.026 | 0.010 | 0.010 | 0.002 |
| 1.5 | 0.016 | 0.195 | 0.167 | 0.026 | 0.008 | 0.007 | 0.422 |
| 2.0 | 0.017 | 0.192 | 0.169 | 0.028 | 0.007 | 0.007 | 0.474 |