tracePlots function

Traceplots for selected parameters

This function draws traceplots of selected parameters from the MCMC simulations ran using NMixMCMC or GLMM_MCMC functions.

tracePlots(x, ...)

  ## Default S3 method:
tracePlots(x, ...)

  ## S3 method for class 'NMixMCMC'
tracePlots(x, param=c("Emix", "SDmix", "Cormix", "K", "w", "mu", "sd", "gammaInv"),
    relabel=FALSE, order,
    auto.layout=TRUE, xlab="Iteration", ylab, col="slateblue", main="", ...)

  ## S3 method for class 'NMixMCMClist'
tracePlots(x, param=c("Emix", "SDmix", "Cormix", "K", "w", "mu", "sd", "gammaInv"),
    relabel=FALSE,
    auto.layout=TRUE, xlab="Iteration", ylab, col=c("blue3", "red3"), main="", ...)

  ## S3 method for class 'GLMM_MCMC'
tracePlots(x, param=c("Deviance", "Cond.Deviance",
    "alpha", "Eb", "SDb", "Corb", "sigma_eps",
    "w_b", "mu_b", "sd_b", "gammaInv_b", "gammaInv_eps"),
    relabel=FALSE, order,
    auto.layout=TRUE, xlab="Iteration", ylab, col="slateblue", main="", ...)

  ## S3 method for class 'GLMM_MCMClist'
tracePlots(x, param=c("Deviance", "Cond.Deviance",
    "alpha", "Eb", "SDb", "Corb", "sigma_eps",
    "w_b", "mu_b", "sd_b", "gammaInv_b", "gammaInv_eps"),
    relabel=FALSE,                             
    auto.layout=TRUE, xlab="Iteration", ylab, col=c("blue3", "red3"), main="", ...)

Arguments

• x: an object of appropriate class.

• param: a character string which specifies which sort of parameters is to be plotted.

  • Emix: overall means (for each margin) of the normal mixture;
  • SDmix: overall standard deviations (for each margin) of the normal mixture;
  • Cormix: overall correlations (each pair) of the normal mixture;
  • K: number of mixture components;
  • w, w_b: weights of each of mixture components. If relabel is FALSE, weights are not re-labeled before plotting;
  • mu, mu_b: component means (each margin, each mixture component) of the normal mixture. The mixture means are shifted and scaled using x$scale$shift and x$scale$scale before plotting. If relabel is FALSE, means are not re-labeled before plotting;
  • sd, sd_b: component standard deviations (each margin, each mixture component) of the normal mixture. The mixture standard deviations are scaled using x$scale$scale before plotting. If relabel is FALSE, standard deviations are not re-labeled before plotting;
  • gammaInv, gammaInv_b, gammaInv_eps: variance hyperparameters;
  • Deviance: deviance (marginal, with random effects integrated out) of the GLMM;
  • Cond.Deviance: conditional deviance (given random effects) of the GLMM;
  • alpha: fixed effects of the fitted GLMM;
  • Eb: overall means (for each margin) of the random effects of the fitted GLMM;
  • SDb: overall standard deviations (for each margin) of the random effects of the fitted GLMM;
  • Corb: overall correlations (each pair) of the distribution of the random effects of the fitted GLMM.
  • sigma_eps: standard deviations of the error terms in the (mixed) models for continuous responses.

• relabel: logical value. It indicates whether the chains with param being w, mu, sd, w_b, mu_b, sd_b should be re-labeled before plotting. Re-labelling is given by argument order. If order is missing then x$order or x$order_b

  determines re-labelling.

• order: a matrix with $K$ columns and $M$ rows where $M$ is the length of MCMC. Each row of order must be a permutation of $(1, ..., K)$.

• auto.layout: logical value. If TRUE, the plotting region is automatically divided to produce traceplots of all parameters. Note that layout must be set up automatically if there are more than 28 parameters to be plotted (often the case for correlations with param being Corb or for mixture means with param being mu_b).

• xlab, ylab, col, main: arguments passed to plot

  function. They all can be of length one (the value is used on all plots) or of length equal to the number of parameters to be plotted.

• ...: other arguments passed to plot function.

Returns

invisible(x)

See Also

NMixMCMC, GLMM_MCMC, NMixRelabel, traceplot.

Author(s)

Arnošt Komárek arnost.komarek@mff.cuni.cz