Effect Size/Mean Squared Error from Linear Mixed-Model in R [closed]

I’m trying to report an effect size for a Linear Mixed-Model we’ve fitted in R. Right now I’m looking at reporting partial eta squared or eta squared. However, to do so I need to calculate the Sums of Squared Error. We’re using the lme() function, which does not report MSE or effect sizes. I am not the one doing the primary analysis, so I don’t know if we can switch to using the ez package, as described in Omega squared for measure of effect in R?.

The lmeObject returned by the lme() function has some information, but I am not sure which is the most appropriate. When using anova() on the lmeObject, it reports the denominator degrees of freedom for the F-test as 56683, so with a value for MSE I can calculate SSE and reverse-engineer the F-tests to get the SStreatments I need for partial eta squared. I have 2 fixed effects and one random effect (for a repeated-measures design).

I’ve looked at lmeObject$sigma and calculated the sums of the squares of lmeObject$residuals[,1], but they don’t agree (I’m squaring the sigma and dividing the SS by the degrees of freedom).

Any R masters out there that can tell how to calculate MSE or SSE from an lmeObject?

Answer

Attribution
Source : Link , Question Author : Oliver , Answer Author : Community

Leave a Comment