Stacked Occu Without Year Random Effect

[Rory Macklin]

Hello all,

Another question. I'm running a stacked occupancy model with two visits to each "site" (site-year combination) in unmarked using the TMB engine. I've attempted to fit these models with Year as a random effect. My data is rather sparse, with few detections - a situation that seems to cause issues for fitting random effects with TMB (https://groups.google.com/g/unmarked/c/7bFNL0BSIx4/m/qZZnYSFyBQAJ). The models will fit, but running GOF tests on them return "system is exactly singular" errors. I am getting errors even when the random effect for year is the only covariate in the models. However, I can fit models with multiple covariates when I include Year as a fixed effect.

Any insight on what I'd be losing out on without the random effect? It seems like the main consequence of stacking is some bias in standard errors, am I correct?

Thanks,

Rory