diagnose.PBNLpostsample function

Diagnostics for the MCMC output in the PB and NL models.

The method issues several convergence diagnostics, in the particular case when the PB or the NL model is used. The code may be easily modified for other angular models.

diagnose(obj, ...)

## S3 method for class 'PBNLpostsample'
diagnose(
  obj,
  true.par = NULL,
  from = NULL,
  to = NULL,
  autocor.max = 0.2,
  default.thin = 50,
  xlim.density = c(-4, 4),
  ylim.density = NULL,
  plot = TRUE,
  predictive = FALSE,
  save = TRUE,
  ...
)

Arguments

• obj: an object of class postsample: posterior sample, as produced by posteriorMCMC.pb or posteriorMCMC.nl
• ...: Additional parameters to be passed to the functions posterior.predictive.pb or posterior.predictive.nl.
• true.par: The true parameter. If NULL, it is considered as unknown.
• from: Integer or NULL. If NULL, the default value is used. Otherwise, should be greater than post.sample$Nbin. Indicates the index where the averaging process should start. Default to post.sample$Nbin +1
• to: Integer or NULL. If NULL, the default value is used. Otherwise, must be lower than Nsim+1. Indicates where the averaging process should stop. Default to post.sample$Nsim.
• autocor.max: The maximum accepted auto-correlation for two successive parameters in the thinned sample.
• default.thin: The default thinning interval if the above condition cannot be satisfied.
• xlim.density: The xlim interval for the density plots, on the transformed scale.
• ylim.density: the ylim intervals for the density plots.
• plot: Logical. Should plots be issued ?
• predictive: Logical. Should the predictive density be plotted ?
• save: Logical: should the result be saved ? Only used if the posterior sample has been saved itself (i.e. if it contains save=TRUE in its arguments list)

Returns

A list made of

• predictive: The posterior predictive, or 0 if predictive=FALSE
• effective.size: the effective sample size of each component
• heidelTest: The first part of the Heidelberger and Welch test (stationarity test). The first row indicates success (1) or rejection(0), the second line shows the number of iterations to be discarded, the third line is the p-value of the test statistic.
• gewekeTest: The test statistics from the Geweke stationarity test.
• gewekeScore: The p-values for the above test statistics
• thin: The thinning interval retained
• correl.max.thin: The maximum auto-correlation for a lag equal to thin
• linked.est.mean: The posterior mean of the transformed parameter (on the real line)
• linked.est.sd: The standard deviation of the transformed parameters
• est.mean: The posterior mean of the original parameters, as they appears in the expression of the likelihood
• sample.sd: the posterior standard deviation of the original parameters