F-Test Calculator

In short An F test compares two variances by dividing one by the other. The samples loaded above have variances of 0.04125 and 0.98125, giving F = 23.79 on 7 and 7 degrees of freedom and a two-tailed p of 0.0005 — the second process is far less consistent than the first, and that is a real finding even though both have similar means. The larger variance always goes on top here, which makes the upper tail the relevant one; putting the smaller on top gives the reciprocal and moves the answer to the other tail, since F(d₁,d₂) = 1 ÷ F(d₂,d₁) with the degrees of freedom swapped. Use this when the SPREAD is the thing you care about: whether one machine is more consistent than another, whether a new process reduces variability, whether two measurement methods agree in precision. What it should not be used for is checking the equal-variance assumption before a t-test. The F test is notoriously fragile on non-normal data — far more so than the t-test it is meant to protect — and reports differences that are not there on anything skewed or heavy-tailed. Levene's test is the robust alternative for that job, and Welch's t-test removes the need to ask at all.