I am wondering what is the difference between multinomial regression and mixed-effects models. When should I apply which of the two algorithms?
Any pointers to literature where the two are discussed would also be highly appreciated.
4
-
1$\begingroup$ Multinomial logistic regression? $\endgroup$– DaveCommented Sep 25, 2020 at 14:58
-
$\begingroup$ edited the question $\endgroup$– leermeesterCommented Sep 25, 2020 at 16:39
-
$\begingroup$ Answer posted, but it would help if you discussed what's confusing you about these ideas, as they're totally different (one dealing with the response variable and one dealing with the predictors). $\endgroup$– DaveCommented Sep 25, 2020 at 16:45
-
$\begingroup$ I'm primarily trying to get a handle on the terminology of different statistical models. As I understand it, statistical models are all variations on y = f(X,b) + error. Where the number of terms on the right hand side might vary, or the link function. I am trying to understand which terminology refers to which variation. $\endgroup$– leermeesterCommented Sep 27, 2020 at 19:43
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
They're completely different notions and even could be combined.
Multinomial logistic regression is for the situation where you want to predict the probability of falling into multiple categories (3+ categories would be multinomial logistic regression...if there are only two categories, it's regular logistic regression).
Mixed effects models are for when your predictor variables include both fixed effects and random effects.
Consequently, if we find ourselves in a position where we have both fixed and random effects as the predictors and want to use them to predict the probability that a photograph is of a dog, cat, or horse, we might use a mixed effects multinomial logistic regression!
