Which family to use in a model?

This seems like a really simple question but I am struggling to find a clear answer.

I am doing an lmer / glmer model.

Can someone please explain or point me in the direction of a clear explanation of a list of families and explanations of when to use each.

Eg. When would you use:
Gaussian
Binomial
Poisson
Quasi
Gamma

What other options are available for lme4 family and when would you use them?
What would the dataset looks like for each would also be useful to see too. Often explanations pick response or explanatory variables but you have no context of what they actually are or what the datasets look like.

A lot of explanations on lmer / glmer just pick a family with no explanation or compare one family over another. I am looking for an overview of all possible families and when to use them that is easy for a novice modeler to understand.

I apologise if this has already come up but I've not come up with anything after spending a morning looking for this online and in books.

Thanks.

These are simply names of probability distrubutions with characteristic shapes/formulations.
E.g. https://www.itl.nist.gov/div898/handbook/eda/section3/eda366.htm
You can look up details all over, i.e. wikipedia, wolframAlpha.

When would you use one or the other... when its a good match with your data's distribution.
I..e you can observe that in a typical population of humans height is distributed roughly normally so gaussian is probably most appropraite

1 Like

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.