blcMCMC.md
In shinyrecap: 'Shiny' User Interface for Multiple Source Capture Recapture Models

MCMC Sampling

These are algorithm convergence controls and should be changed in response to issues raised by the model diagnostics.

| Problem | Solution | ------ | -------- | Too low an effective sample size (e.g. <500) and or the trace plot shows auto correlation. | Increase either # of samples or Thinning | | There is a trend at the beginning of the trace plot | Increase Burn In | | Algorithm is taking too long to run | Decrease # of samples or Thinning |

Definitions: # of samples: The number of MCMC samples to draw from the posterior. Thinning: Only sample every nth realization from the MCMC chain. * Burn in: Drop the first n realizations from the MCMC chain.