stdranef: Extract standard deviation of "random" effects from an...

In JWiley/postMCMCglmm: Average marginal predicted probabilities from Bayesian ordered probit models

Description

Function designed to extract the standard deviation of the random effects from an MCMCglmm model object. Note that this is not the same as the posterior distribution of (co)variance matrices. It is based on the posterior distribution of the random effects. This also means it requires pr=TRUE to be set in the model for the information to be saved. Can optionally return standard deviation of random effects after back transforming to the response metric. Currently probabilities, but only for ordinal family models (family="ordinal").

Usage

  stdranef(object, which, type = c("lp", "response"), ...)

Arguments

object	An MCMCglmm model object to extract the effects from
which	A list of random effects to extract or their numeric positions If there are two numbers in a list, effects are simulataneous.
type	A chacter string indicating whether to calculate the standard deviation on the linear predictor metric, ‘lp’ or response, ‘response’.
...	Not currently used.

Value

A list of class postMCMCglmmRE with means (M) and individual estimates (Data)

See Also

print.postMCMCglmmRE, predict2.MCMCglmm, ranef.MCMCglmm

Examples

## Not run: 
  # a simple MCMCglmm model
  data(PlodiaPO)
  PlodiaPO <- within(PlodiaPO, {
    PO2 <- cut(PO, quantile(PO, c(0, .33, .66, 1)))
    plate <- factor(plate)
  })

  m <- MCMCglmm(PO2 ~ 1, random = ~ FSfamily + plate,
    family = "ordinal", data = PlodiaPO,
    prior = list(
      R = list(V = 1, fix = 1),
      G = list(
        G1 = list(V = 1, nu = .002),
        G2 = list(V = 1, nu = .002)
      )
    ), verbose=FALSE, thin=1, pr=TRUE)

  # summary of the model
  summary(m)

  # examples of extracting standard deviations of
  # different random effects on the linear predictor metric
  # or after transformation to probabilities (only for ordinal)
  stdranef(m, which = list(1), type = "lp")
  stdranef(m, which = list(2), type = "lp")
  stdranef(m, which = list(1, 2, c(1, 2)), type = "lp")
  stdranef(m, type = "lp")

  ## error because no 3rd random effect
  #stdranef(m, which = list(1, 2, 3), type = "lp")

  stdranef(m, which = list("FSfamily", "plate"), type = "lp")

  # mean standard deviations on the probability metric
  # also the full distributions, if desired in the Data slot.
  res <- stdranef(m, type = "response")
  res$M # means
  hist(res$Data$FSfamily[, 1]) # histogram

## End(Not run)

JWiley/postMCMCglmm documentation built on May 7, 2019, 10:15 a.m.