search

pitakakariki / simr

Power Analysis of Generalised Linear Mixed Models by Simulation
70 stars 19 forks source link

Setting effect size #111

Closed sho-87 closed 6 years ago

sho-87 commented 6 years ago

A couple of questions about setting fixed effect sizes:

  1. I'm using fixef(model)["varA"] <- -0.1 to set the fixed effect for one of my predictors. Should this be done for all predictors in the model? Or only the one of interest?

  2. What metric is the effect size in? I'd like to set a standardized effect size like a Cohen's d, but even a value like 0.1, which is pretty small in cohen's d terms, gives me a lot of power when looking at the power curve

pitakakariki commented 6 years ago

  1. This is probably best addressed here: https://github.com/pitakakariki/simr/issues/112

  2. As far as I know, there's no standard equivalent to Cohen's d for mixed models. The effect size here is the same as given by fixef, so whether it's small or large depends on what scale your covariates are on.