GLMM predictor variables I have a question regarding predictor variables in a GLMM analysis. 
I am running a study investigating multi-modal communication in primates,
specifically, primate gestures that accompany vocalizations. 
I have measured the call rate, call duration and peak frequency (response variables) for 200 calls (100 accompanied by gestures, 100 without gestures). 
I would like to use the following predictor variables: 1) Call type, 2) Presence of gesture, 3) Gesture type. 
Is it possible to use GESTURE TYPE here as a predictor variable? I am not sure if it's appropriate since this will only be for 50% of the overall the data (where calls are accompanied by gestures). 
 A: Expanding on  @Nick 's correct answer, I think you have three choices:
1) You could code "gesture" as a single indicator variable (I had not heard that objection to 'dummy')
2) You could code each particular gesture type as an indicator variable, and if all of them are 0 then there is no gesture at all.
3) You could run two analyses, one for communication where there is a gesture and one where there is not.
Of these, 1) is completely standard but may not answer your question. 3) is also fine if there is enough data to split the data set. 2) is a little less standard in my experience, but I do not see a problem with it.
A: Gesture type could be coded 0-1 for without-with (or vice versa) and that could be a predictor. Such a predictor is often called a dummy or indicator variable. (One objection to the term "dummy" is that it has been been misread as offensive or dismissive by people who don't realise it is arbitrary mathematical jargon. I have real stories on this, but the details are best suppressed in this forum.) 
Whatever you call it, using such a binary or dichotomous predictor (yet more terms) is an utterly standard method. Whether it is a good idea in your case depends on your data generating process. 
