mi_lm_pai_cvboot: Calculate PAI on multiply imputed data in parallel

In JWiley/pai: Calculate bootstrapped and cross-validated personalized advantage index

Description

This is a simple function that wraps lm_pai_cvboot.

Usage

mi_lm_pai_cvboot(formula, DV, TreatVar, data, nboot = 500, holdouts, cl,
  cores, seed = 1234, boot = TRUE, yhigherisbetter = TRUE)

Arguments

formula	An R formula, but without dependent variable
DV	a character string indicating the name of the dependent variable
TreatVar	a character string indicating the name of the treatment variable (note that values must be coded as 0/1 for the treatment variable.
data	A list of imputed datasets or for a single dataset, a list of one
nboot	The number of bootstrap resamples to take
holdouts	The holdouts to be used. This should be a numeric list with as many elements desired for holdouts. For example, if the list is the same length as the number of cases and each element is one case (i.e., 1, 2, ..., N), then it is equivalent to leave-one-out. If there are ten elements, each containing approximately 1/10th of the sample, it is 10-fold cross validation, etc. If a single character string is passed, K-fold CV is assumed and holdouts are automatically constructed.
cl	An existing cluster to use (optional)
cores	The number of cores to use when creating a cluster, if an existing cluster is not passed (optional). If left blank and no cluster passed, defaults to the number of cores available.
seed	A seed to make the bootstrap reproducible
boot	A logical value whether to bootstrap or not
yhigherisbetter	A logical value whether higher scores indicate better treatment response on the dependent variable

Value

A list containing a four dimensional array of the results, as well as the input parameters for the model. For the array, the first dimension is cases, the second dimension is multiple estimates calculated, the third dimension is bootstrap resamples, and the fourth dimension is imputations.

Examples

## Not run: 
# builtin dataset with 32 cases
dat <- mtcars
dat$cyl <- factor(dat$cyl)
formula <- ~ cyl + am * (mpg + hp + drat + wt)
DV <- "disp"
TreatVar <- "am"
k <- 50 # use a few just for speed
holdouts <- 1:32 # for leave-one-out

m <- mi_lm_pai_cvboot(formula, DV, TreatVar, list(dat, dat),
  nboot = k, holdouts = holdouts, cores = 2)

## End(Not run)

JWiley/pai documentation built on May 7, 2019, 10:15 a.m.