I have ecological data that is spread over three years. My data looks at repeated observations through time (0.43, 1, 2, and 3 weeks). The data is unbalanced, not all levels of week have observations. I would like to examine how insect communities change from one level of ‘week’ to the next, so I made ‘week’ an ordered factor. I wanted to do this while accounting for ‘year’ and ‘seasonality’. My data is set up so each row represents a unique observation and taxa.

I tried glmer.nb and some zero-inflated models, generally with glmer’s they fail to converge because my data has many zeros and I have probably around 400 observations, 3 years, 4 levels of week, and data covers 4-5 months for seasonality. I am wondering if I should look at random forest analysis or something along those lines? Thanks, M