Analyzing animal behavioural data in R Studio (GLMM) and struggling to understand how to set up the model I'm trying to determine whether there was any differences in behaviour (count data, frequency per 30 minute observations, categorized into 5 categories [e.g., affiliative, agonistic, etc.]) of a captive troop of baboons between lockdown (no visitors) and out of lockdown (varying numbers of visitors). I also have environmental variables data (e.g., weather, temperature, wind speed, etc.) and management variables (e.g., group size present, time since last feed, etc.).
I don't know if I have too many random effects for one generalized linear mixed model (GLMM) or if that's even the correct test to do. I've been working with the following code:
Affiliative <- glmer(Affiliative/Total ~ Feed.Lag + Weather + Temperature + Group.Size + (1 | Open), data = baboon, family = binomial, na.action = na.omit, weight = Total)

I've set Weather and Open (i.e., lockdown vs. no lockdown) to as.factor() variables, as they are non-numeric and just variables given an arbitrary number for my analysis. The rest of the variables I've left alone as they are numeric.
I thought I'd use each behavioural category one at a time to see which variable, if any, changed behaviour, but I'm unsure if I've set up the code up correctly. It runs and I get significant results back, but my understanding of which variables are fixed/random and where they fit in the code isn't great.
Any help really appreciated, thank you!
 A: Your model:
Affiliative <- glmer(Affiliative/Total ~ Feed.Lag + Weather + Temperature + Group.Size + (1 | Open), data = baboon, family = binomial, na.action = na.omit, weight = Total)

does not make sense.
Since Open is a factor with 2 levels you cannot use it as a grouping variable for random intercepts. When you fit random intercepts you are telling the software that this is a normally distributed variable and you are asking it to estimate the variance of it. With only 2 observations, this is not sensible. It should be a fixed effect.
From the information given I don't see any reason to fit a mixed effects model. A GLM (either poisson or binomial) would be more appropriate.
In the comments you say:

I know that once it's running properly I can remove those that aren't significant

Please do not do this. Variables should be included in a model because the underlying theory dictates so, not because a p value is below some arbitrary value. Variables should be included when they are competing exposures or potential confounders. If any variable is a mediator it should be omitted.
