calculating degrees of freedom mixed factor anova

Calculating degrees of freedom mixed factor anova code#

– the nice thing here is that you can also read the article’s peer review history (see the rightmost column). “Should I use fixed effects or random effects when I have fewer than five levels of a grouping factor in a mixed-effects model?”. But there are a few papers that you might find useful (and the references therein, of course): When I last looked into this topic (~Fall 2021), most of what I found was specific to ecological research. The second reference below ran simulations with as low as 2 levels.

Calculating degrees of freedom mixed factor anova code#

Regarding second sentence, it looks like it is mathematically possible, or at least the code will execute (perhaps the function would convert a random effect with 2 levels into a fixed effect) but I’ve never tried it. So, one of your models in which time is not a random effect would work (assuming id refers to patients, and you have a sufficient number of patients for this to be sensible). However, the accepted answer links to Ben Bolker’s GLMM page which suggests at least 5 or 6 levels for a given random effect, and the next highest answer quotes Gelman & Hill as saying a random effect with fewer than 5 levels won’t add much over a standard LM. It appears (for this example, anyway) that the variance doesn’t really stabilize until there are at least 14 subjects, if not later.

To this point specifically (the first sentence), see this answer on Stack Exchange for some simulation results for a LMM. It may even be mathematically impossible to fit such a model. I don’t think random effects work well with only 2 time points.