logLik.merr

Observed-data log-likelihood for a merr object via Monte Carlo integration over the measurement error distribution. Enables AIC(fit) and BIC(fit) automatically.

Usage

# S3 method for class 'merr'
logLik(object, ndraws = 500, ...)

Arguments

object

A merr object returned by qrme().

ndraws

Number of Monte Carlo draws for the integral (default 500).

...

Currently unused.

Value

A scalar of class "logLik" with attributes df (number of free parameters) and nobs (sample size).

Examples

# \donttest{
  set.seed(42)
  n <- 200; X <- runif(n)
  Y <- 1 + 2 * X + rnorm(n) + rnorm(n, sd = 0.5)
  fit1 <- qrme(Y ~ X, data.frame(Y, X), tau = 0.5,
               n_mix = 1L, mcmc_draws = 80L, mcmc_burn_in = 40L,
               max_em_iters = 10L, verbose = FALSE)
  fit2 <- qrme(Y ~ X, data.frame(Y, X), tau = 0.5,
               n_mix = 2L, mcmc_draws = 80L, mcmc_burn_in = 40L,
               max_em_iters = 10L, verbose = FALSE)
#> Warning: Too many fixups:  doubling m
#> Warning: EM algorithm failed to converge after 10 iterations
  data.frame(n_mix = 1:2,
             AIC   = c(AIC(fit1), AIC(fit2)),
             BIC   = c(BIC(fit1), BIC(fit2)))
#>   n_mix      AIC      BIC
#> 1     1 745.0586 756.4428
#> 2     2 746.3905 766.1827
# }