In the analysis of microarray data one generally produces a vector of p-values that for each gene give the likelihood of obtaining equally strong evidence of change by pure chance. The distribution of these p-values is a mixture of two components corresponding to the changed genes and the unchanged ones. The basic question 'What proportion of genes is changed' is a non-trivial one, with implications for the way that such experiments are analysed. An estimate not requiring any assumptions on the distributions is proposed and evaluated. The approach relies on the concept of a moment generating function.
A simulation model of real microarray data was used to assess the proposed method. The method fared very well, and gave evidence of low bias and very low variance.
The approach opens up a new possibility of sharpening the inference concerning microarray experiments, including more stable estimates of the false discovery rate.