stackMI: Bayesian Model Fitting and Stacking for Multiple Imputation
In rmsb: Bayesian Regression Modeling Strategies

stackMI

R Documentation

Bayesian Model Fitting and Stacking for Multiple Imputation

Description

Runs an rmsb package Bayesian fitting function such as blrm separately for each completed dataset given a multiple imputation result such as one produced by Hmisc::aregImpute. Stacks the posterior draws and diagnostics across all imputations, and computes parameter summaries on the stacked posterior draws.

Usage

stackMI(
  formula,
  fitter,
  xtrans,
  data = NULL,
  n.impute = xtrans$n.impute,
  dtrans = NULL,
  derived = NULL,
  subset = NULL,
  refresh = 0,
  progress = if (refresh > 0) "stan-progress.txt" else "",
  file = NULL,
  ...
)

Arguments

`formula`	a model formula
`fitter`	a Bayesian fitter
`xtrans`	an object created by `transcan`, `aregImpute`, or `mice`
`data`	data frame
`n.impute`	number of imputations to run, default is the number saved in `xtrans`
`dtrans`	see `Hmisc::fit.mult.impute`
`derived`	see `Hmisc::fit.mult.impute`
`subset`	an integer or logical vector specifying the subset of observations to fit
`refresh`	see rstan::sampling. The default is 0, indicating that no progress notes are output. If `refresh > 0` and `progress` is not `''`, progress output will be appended to file `progress`. The default file name is `'stan-progress.txt'`.
`progress`	see `refresh`. Defaults to `''` if `refresh = 0`. Note: If running interactively but not under RStudio, `rstan` will open a browser window for monitoring progress.
`file`	optional file name in which to store results in RDS format. If `file` is given and it already exists, and none of the arguments to `stackMI` have changed since that fit, the fit object from `file` is immediately returned. So if the model, data, and imputations have not changed nothing needs to be computed.
`...`	arguments passed to `fitter`