combine.AIC: Combining AICs

In miWQS: Multiple Imputation Using Weighted Quantile Sum Regression

Description

Combines individual AIC estimates of separate models to get a sense of overall model fit.

Usage

combine.AIC(AIC)

Arguments

AIC	A vector of AICs to combine with length equal to the number of models completed (i.e. K).

Details

The WQS model fits using different completely observed datasets are combined in Stage 3 of multiple imputation. Similar to combining WQS parameter estimates, the mean of individual AIC estimates is taken as the central tendency estimate of WQS model fit. The standard deviation between individual AIC estimates indicates the difference in WQS model fits due to below the detection limit values.

A vector of AICs may be generated from do.many.wqs().

Value

The overall fit of a model across all imputation models: the mean AIC +/- the standard error. Saved as a 1x1 character vector.

Warning

If AIC is a vector with one element, the AIC is returned as a character rounded to the nearest whole number with a warning printed that AIC cannot be combined.

See Also

pool.mi

Examples

# AICs from do.many.wqs() example are as follows.
bayes.AIC <- c(1295.380, 1295.669)
combine.AIC(bayes.AIC)

# One AIC
combine.AIC(1295.380)

miWQS documentation built on April 3, 2021, 1:06 a.m.