AndersonAll the best,Hi,There's no strict rule for this, but the best is that the value isn't too low. Say your F-test has df1=2 and df2=100. An F-statistic that produces p=0.001 is 7.4, and this could be used. But the 0.001 is arbitrary, and could as well have been another value. It could indeed be 3 (that gives p~0.05 with similar df).