I am trying to fit a mixed model GLM for circular data on R (circular response variable, theta (I have tried this for radians and degrees) and continuous and/or categorical predictor variables. I have run this with standardised predictor values and raw predictor values and run into the same error. An example of the model is below, based on Cremers 2018:
p19.fit <- bpnme(pred.I = theta ~ class + native.stsd + julian.stsd + (1|index), data = puma.glmm, its=1000, burn=00, n.lag=3)
where class is a categorical variable with 2 levels, native.stsd is standardised native vegetation percentage, and julian.stsd is standardised julian day, and index is the random variable. I have tried running traceplots for non-standardised values, for different combinations of other predictors (including only categorical or only continuous, etc.). I have also tried dropping the index variable and running a normal GLM (using bpnr) and finally I have tried numerous combinations of the value for iterations, burn, and n.lag.
No matter what combinations I do, I appear to have traceplots that are not converging and the results associated with the models seem to be skewed (large standard deviations, no results where results are expected and vice versa). I'm wondering if there is an issue with my code/data?
