Modeling Advice

This is a collection of advice for modeling eDNA data with the artemis package.

  1. Center and scale your predictor values: artemis uses MCMC to estimate values, and this will be more efficient if the predictor values are not on vastly different scales. In general, the MCMC will be the most efficient when the predictors are roughly centered at 0, and have sd of 1.

  2. Use priors: The default priors in artemis follow the conventions of the rstanarm package, and are weakly informative. When the data do not strongly inform the parameter estimates, the model fit can be improved by specifying stronger priors.