warning epsilon > 1 for HF correction

[statmerkur]

I am getting the following warning message when running a repeated measures anova using aov_ez:
In summary.Anova.mlm(object$Anova, multivariate = FALSE) :
HF eps > 1 treated as 1
What does it mean?

[henrik]

This warning hopefully only appears if you actually use HF as correction. It indicates that the Huynh-Feldt epsilon is above 1. This is the correction factor applied to the degrees of freedom for violations of sphericity. Values above 1 do not make any sense, as the df are multiplied with this correction factor and should not be above the uncorrected df (which would appear for values above 1). Thus, for epsilons above 1, 1 is used. See also: https://en.wikipedia.org/wiki/Mauchly%27s_sphericity_test

You did not include a reproducible example, so I am not sure if this appears incorrectly. In any case, this is not a critical warning and simply passed on from summary.Anova.mlm which produces it.

[statmerkur]

Yes, it only appears if I use the HF. I didn’t know that epsilon can be above 1 – thanks for the hint.

[Author]

Posts