Figure 3. Comparison of the p-value and z-score distributions for all four methods.

The analysis were performed on a set of randomized KEGG pathways. All results were filtered for reduced chi-squared < = 1.0. According to the central limit theorem the distribution of z-scores between random groups should be standard normal and therefore give a uniform distribution of p-values. For each p-value distribution we report the ratio between the number of p-values <0.05 and the average number in equally sized bins with p-value > = 0.05. An unbiased distribution should give a ratio of 1. For each z-score distribution we report skewness, mean and standard deviation. For each method 150 rando­mizations were performed.