I am having trouble fitting a mixed effect zero inflated negative binomial model to my data using the GLMMadaptive package:
negbi_1 <- mixed_model(fixed=MA ~ ST + AG + SU +SO +Y, random = ~ 1|FA, data = compressed_noprice_isolatednew, family=zi.negative.binomial(), zi_fixed = ~ SE)
returns: Error in chol.default(X[[i]], ...) : the leading minor of order 1 is not positive definite I can't find anything about this particular error message for these models.
My data has 946 rows. The resonse MA is bird survey count data (MA). All of my predictor variables are categories; ST indicates the protection status of a bird (3 levels), Age represents Age (2 levels, adult and juvenile) and SU represents survey timing (4 levels)
SE (4 levels) indicates the season the survey took place in, which could reduce the abundance value, often to 1, which is why I am trying to use a zero-inflated model.
I was using this tutorial https://drizopoulos.github.io/GLMMadaptive/articles/Goodness_of_Fit.html?fbclid=IwAR2D4KOC61b7xivHj32uuvd7KhHbaTaMWjoiRgtmplzb1dY2EJWPfFzqLFs, though I can not start even building the model.
I also tried all the options from this tutorial: and I still get the error codes, Mixed effect zero inflated negative binomial model: "the leading minor of order 1 is not positive definite"
I have also tried a negative binomial distribution in glmmTMB and in lme4 using the glmer.nb function did not produce a significant qq plot, it seemed slightly underdispersed. I tried to use k-inflation methods, as I have many 1's. I have tried so many different combinations of models. I can explain further if the wording is not cler, I just can't find any answers anywhere.
I have also not scaled and centred the dataset and am not sure how to do this or if its possible with categorical variables.
I also looked at this vignette- https://cran.r-project.org/web/packages/glmmTMB/vignettes/model_evaluation.pdf
I think the high number of observations with few individuals is causing problems in model adjustment? Perhaps I need to use a truncated model, as shown here: https://github.com/florianhartig/DHARMa/issues/131 OR https://www.biorxiv.org/content/biorxiv/suppl/2017/05/01/132753.DC1/132753-2.pdf
