bottle: Bottle a brew

In bcjaeger/ipa: Imputation for Predictive Analytics

Description

Bottle a brew

Usage

bottle(brew, type = "tibble", drop_fit = TRUE)

Arguments

brew	an ipa_brew object.
type	a character value indicating what composition the training and testing data should have. Valid options are 'tibble' and 'matrix'
drop_fit	a logical value. If TRUE, the column in the wort comprising imputation models will be dropped. Otherwise, the column will be retained.

Examples

x1 = rnorm(100)
x2 = rnorm(100) + x1
x3 = rnorm(100) + x1 + x2

outcome = 0.5 * (x1 - x2 + x3)

n_miss = 10
x1[1:n_miss] <- NA

data <- data.frame(x1=x1, x2=x2, x3=x3, outcome=outcome)

sft_brew <- brew_soft(data, outcome=outcome, bind_miss = FALSE)
sft_brew <- mash(sft_brew, with = masher_soft(bs = TRUE))
sft_brew <- stir(sft_brew, timer = TRUE)

data_new = data.frame(
  x1      = c(1/2, NA_real_),
  x2      = c(NA_real_, 2/3),
  x3      = c(5/2, 2/3),
  outcome = c(1/3, 2/3)
)

# soft models are re-fitted after stacking data_new with data_ref

sft_brew <- ferment(sft_brew, data_new = data_new)

bottle(sft_brew, type = 'tibble')

bcjaeger/ipa documentation built on May 7, 2020, 9:45 a.m.