ESS greater than total posterior samples

[Marc Kéry]

Dear Jeff,

happily continuing my exploration of static multispecies occupancy models with spOccupancy here. With 105 species at 2154 sites and >30 covariates in occupancy, I get a nonspatial model to convergence (for most params) in about 3 hours on my sluggish laptop (which, moreover, churns away on another big spOccupancy job).

I just discovered that the effective sample size ESS can be greater than the nominal sample size. I keep only 100 posterior draws from each parameter, but some values of ESS are far greater, going up to 400. I haven't checked how ESS is computed, but assume that this is fine ?

Thanks and best regards  --- Marc

[Jeffrey Doser]

Hi again Marc,

The package calculates effective sample size using coda::effectiveSize. A larger value of ESS than the actual sample size I believe means that there is negative autocorrelation in the estimates, which I don't think should be a problem as far as quantifying uncertainty goes. I do know that there are multiple ways to calculate ESS beyond just the approach that is used by coda, but I am not super familiar with knowing the pros and cons to the different approaches.

Jeff

--
You received this message because you are subscribed to the Google Groups "spOccupancy and spAbundance users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to spocc-spabund-u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/spocc-spabund-users/CAP5oAEr_eUt_J0cUiFSyKXn2e%2BNRswuMAj1V4HbxvV8vOqzzRA%40mail.gmail.com.

--

Jeffrey W. Doser, Ph.D.

Assistant Professor

Department of Forestry and Environmental Resources

North Carolina State University

Statistical Ecology and Forest Science Lab

Pronouns: he/him/his