How to do GLMM if the dependent variable is categorical in R? I'm trying to run a logistic regression on categorical dependent variables data; The dependent variable is a choice of Yes/No/Unsure.
I want to use glmer in lme4, how do I specify the family type?
If I transformed the dependent variables into rate of Yes (ACC), the distribution of ACC look like the histogram below. Can I use the following formula to fit the model?
model <- lmer(ACC ~ group * tone * focus + (1|subject/word), data = d)

Fixed effects: group, tone, focus
Random effects: subject, word

 A: As far as I know, there isn't a way to conduct a multinomial logistic regression using the lme4 package. I have this post from SO saved for when I am in the situation you are in. If I am wrong, I would love for someone to correct me, because I would like to use that functionality.
When I have 3 levels of a categorical dependent variable, I have fit three separate models before: 


*

*DV in first model: Level 1 or Levels 2 and 3

*DV in second model: Level 2 or Levels 1 and 3

*DV in third model: Level 3 or Level 1 and 2
It seems like you are OK with collapsing across Levels 2 and 3 (No and Unsure). So what I would do is recode your dependent variable to be dichotomous: 
d$newDV <- as.factor(ifelse(d$oldDV=="Yes", 1, 0))

Then you could run a model like:
model <- glmer(newDV ~ group * tone * focus + (1|subject/word), 
               data = d, family=binomial)

Keep in mind, though, that you will want to likely include random slopes for group, tone, and focus (or at least test to see if they are significant). The code would look something like this, but only if all of the three predictors are at the lowest level in the multilevel structure of the data:
(1+group+tone+focus|subject/word)
