Error variance is defined as a variance computed to measure the magnitude of differences that would be expected if the null hypothesis is true and there are no population mean differences. The denominator of the F-ratio computed in an analysis of variance.