Objectivity of the measure of relevance.

Objectivity of the measure of relevance for the two sample-problem expressed in sensitivity and specificity indexes. Three two-sample situations were simulated. In each of them 1,000 variables among which 10 were very informative (normalized biological difference (NBD)), semi-informative (NBD) or non-informative (NBD), respectively. With simulation runs the sensitivity (% correctly identified informative variables) and specificity (% correctly identified non-informative variables) were determined. Specificity is excellent and with increased normalized biological differences and sample sizes also the sensitivity indexes go up to excellent levels 100% (for more details see text).aRates (in %) of variables detected as relevant by the .bRates (in %) of variables detected as significant by applying appropriate nonparametric tests.