latent_operation: Operation to conduct in the latent space

In udellgroup/mixedgcImp: Missing value imputation for mixed data through Gaussian copula

Operation to conduct in the latent space

Description

Conduct one of three tasks in the latent space. If task = "em", conduct an em iteration. If task = "fillup", impute the missing entries of Z as their conditional mean. If task = "sample", conduct multiple imputation on the missing entries of Z.

Usage

latent_operation(
  task,
  Z,
  Lower,
  Upper,
  d_index,
  dcat_index,
  cat_input,
  corr,
  trunc_method = "Iterative",
  n_update = 1,
  n_sample = 5000,
  n_MI = 1
)

Arguments

task	Task to perform. One of "em", "fillup", "sample".
Z	Transformed latent matrix
Lower	Lower boundary of truncated intervals
Upper	Upper boundary of truncated intervals
d_index	Boolean vector with TRUE at ordinal dimensions
dcat_index	Boolean vector with TRUE at categorical dimensions
cat_input	Input for categorical dimensions
corr	Current copula correlation estimate
trunc_method	Method for evaluating truncated normal moments: 'Iterative' or 'Sampling'.
n_update	The number of updates, only used when trunc_method is 'Iterative'
n_sample	Number of MC samples, only used when trunc_method is 'Sampling'

Value

A list containing

corr

Available when task = 'em'. Updated correlation estimate.

loglik

Available when task = 'em'. The average log-likelihood.

Z

Available when task = 'em'. Incomplete Z with updated observed ordinal entries

Zimp

Available when task = 'em' or task == 'fillup' . Complete Z with observed entries the same as Z and missing entries imputed

Zimp_sample

Available when task = 'sample'. Multiple imputation samples.

C

Available when task = 'em'. The conditional co-variance due to missingness

var_ordinal

Available when task = 'em' or task = 'fillup'. The conditional variance due to truncation, i.e. Var(z|a < z < b)

udellgroup/mixedgcImp documentation built on Jan. 25, 2023, 7:55 p.m.