plotBUGSTrace: Trace Plots of MCMC Chains from BUGS
In aquaMetrics/fcs2: Fisheries Classification Scheme 2 For SNIFFER

Description Usage Arguments Details See Also

Produces trace plots of the MCMC samples for each parameter in the fitted FCS2 model. These can be used to visualise the convergence of the MCMC chains and the autocorrelation.

plotBUGSTrace(
  fit,
  variables = variable.names(fit, rw = "singular", spatial = "singular", hyperpar =
    "prec"),
  chains = 1:fit$bugsFit$n.chains,
  nPerPage = 12,
  col = chains + 1,
  ylab,
  ylim,
  main,
  ...
)

`fit`	an `"fcs2Fit"` object containing a full FCS2 model fit, as returned from `fcs2FitModel` with `runBUGS = TRUE`.
`variables`	an optional character vector giving the names of the model variables to plot. `fcs2:::variable.names.fcs2Fit` can be used for this, but only singular variable names can be given.
`chains`	which MCMC chains to plot. Default to all.
`nPerPage`	the number of trace plots to create on each page.
`col`	a vector giving the colour to use for each chain in `chains`.
`ylab, ylim, main, ...`	further arguments passed to `plot` when creating each plot.

If the MCMC chains have converged, the multiple chains (in different colours) should agree; Their values should not be identical but they should cover the same range of values as each other.

The MCMC chains should also mix well meaning that they should move about the parameter space quickly. A chain that moves slowly across the parameter space has high autocorrelation. This can also be verified by plotting the autocorrelation of a parameter using acf. The autocorrelation can be reduced by thinning the sample using thinBUGSSamples or by specifying a larger number for n.thin when fitting.

thinBUGSSamples

aquaMetrics/fcs2 documentation built on Aug. 21, 2021, 12:55 p.m.