Improved estimates for the CPO/PIT-values

Improve the estimates of the CPO/PIT-values be recomputing the model-fit by removing data-points.

inla.cpo(
  result,
  force = FALSE,
  mc.cores = NULL,
  verbose = TRUE,
  recompute.mode = TRUE
)

Arguments

result

An object of class inla, ie a result of a call to inla()

force

If TRUE, then recompute all CPO/PIT values and not just those with result$cpo$failure > 0.

mc.cores

The number of cores to use in parallel::mclapply. If is.null(mc.cores), then check getOption("mc.cores") and inla.getOption("num.threads") in that order.

verbose

Run in verbose mode?

recompute.mode

Should be mode (and the integration points) be recomputed when a data-point is removed or not?

Value

The object returned is the same as result but the new improved estimates of the CPO/PIT values replaced.

See also

inla()

Author

Havard Rue hrue@r-inla.org

Examples

n = 10
y = rnorm(n)
r = inla(
  y ~ 1,
  data = data.frame(y),
  control.compute = list(cpo=TRUE),
  num.threads = "1:1" # Protect package testing from parallel execution
)
# \donttest{
rr = inla.cpo(
  r,
  force=TRUE,
  mc.cores = 1 # Protect package testing from parallel execution
)
#> Compute new CPO/PIT values manually, for 10 cases...
#>       index   cpo.old   cpo.new  pit.old  pit.new
#>  [1,]     1 0.2293658 0.2293658 0.648756 0.648756
#>  [2,]     2 0.1977867 0.1977867 0.256690 0.256690
#>  [3,]     3 0.1300788 0.1300788 0.133419 0.133419
#>  [4,]     4 0.1463859 0.1463859 0.158112 0.158112
#>  [5,]     5 0.0498728 0.0498728 0.961189 0.961189
#>  [6,]     6 0.1691352 0.1691352 0.197014 0.197014
#>  [7,]     7 0.0118922 0.0118922 0.992734 0.992734
#>  [8,]     8 0.1858985 0.1858985 0.230086 0.230086
#>  [9,]     9 0.2473849 0.2473849 0.522305 0.522305
#> [10,]    10 0.2463437 0.2463437 0.541909 0.541909
#> 
#> The retured result contain the new values.
# }