nlsymiss dataset (581 children from the NLSY with missing
values on five variables), used in all examples.nlsymiss instead of being wrapped in
\dontrun{}.glemb() with an odd number of
imputations (m) caused results to alternate between two distinct sets of
imputations across successive calls with the same seed. The imputation step
now uses R's own random number generator instead of the Fortran RNG in the
mix package, whose internal state could not be fully reset between calls.max.cells argument to glemb() (default 500000). A pre-flight
check now stops with an informative error before glemb allocates any memory
when the nominal variables would produce more cells than this limit. Set
max.cells = Inf to disable the check.meanmodel argument to glemb(). "main" (default) constrains the
continuous variable means to depend only on the main effects of the
categorical variables, reducing the number of mean parameters from
n_cells to 1 + sum(K_j - 1). "saturated" restores the original
cell-specific intercepts (unrestricted means).meanmodel value is stored in the returned object and displayed by
print() and summary().output = "mitml" option, returning a mitml.list object compatible
with mitml::testEstimates().Initial release.
glemb(): multiple imputation for mixed continuous and categorical data
using the General Location Model with the EMB algorithm.cat.interact = 3.cat.prior) for sparse-cell
stabilisation.empri).mids object
compatible with mice::pool().print() and summary() methods for glemb objects.