probitCJS Error: vcv matrix not avaiable in model

[emilisak]

I have a dataset of 300+ individuals, each with a capture history of length 60. For Phi, I have variables: sex (binary factor), origin (binary factor), days between capture occasion and first recapture (continuous numerical), season (time-varying 4-level factor), and age (time-varying numeric). For p, I have sex, origin, days between capture occasion and first recapture, trap-dependency (as described in the marked vignette), days per season we tried to recapture individuals (continuous numerical), and season (time-varying 4-level factor).

If I run a full parametrized model with crm(...), it works fine. However, when I run a probitCJS model, I get the error that "vcv matrix not available in model". And if I run it with 500K iterations (reasonable iteration based on similar studies), I get error "cannot allocate vector of size ~50GB.
Regardless of MCMC specifications, any probitCJS model I try to run gives warnings that the functions 'nobars' and 'findbars' have moved to the formulas package.

I've tried to find explanations for these issues elsewhere without luck.
What do these errors entail and is there a fix to them?
Thank you for the help.

[dsjohnson]

I'll dig into the code and check it out with respect to the vcv error and the 'bars' issues. However, there may not be much that can be done with the allocation error. It sounds like the storage it needs to allocate in order to save that amount of data is too big for R. If you really want to use MCMC for this analysis you might have to switch to a more efficient method, e.g., coding this up in nimble or jags. Does this same allocation error happen if you specify, say, 50,000 iterations?
-Devin