sip: Sip a brew

In bcjaeger/ipa: Imputation for Predictive Analytics

Description

Sip a brew

Usage

sip(
  brew,
  from = "training",
  data_complete,
  fun_ctns_error = yardstick::rsq_vec,
  fun_intg_error = yardstick::rsq_vec,
  fun_bnry_error = yardstick::kap_vec,
  fun_catg_error = yardstick::kap_vec
)

Arguments

brew	an ipa_brew object.
from	column name for sipping. Valid options are training and testing. Inputs can be quoted (e.g. 'training') or unquoted (e.g. training).
data_complete	a data frame containing the 'true' values that were 'missing'.
fun_ctns_error	a function that will evaluate errors for continuous variables. Continuous variables have type double. Default is to use R-squared (see yardstick::rsq()).
fun_intg_error	a function that will evaluate errors for integer valued variables. Default is to use R-squared (see yardstick::rsq()).
fun_bnry_error	a function that will evaluate errors for binary variables (i.e., factors with 2 levels). Default is to use kappa agreement (see yardstick::kap()).
fun_catg_error	a function that will evaluate errors for categorical variables (i.e., factors with >2 levels). Default is to use kappa agreement (see yardstick::kap()).

Value

an ipa_brew object with a new column added to the wort. The new column contains a list of tibble::tibble()s with columns variable, type, and score. The score column comprises output from the error functions.

Examples

data("diabetes")

df_miss <- diabetes$missing
df_cplt <- diabetes$complete

sft_brew <- brew_soft(df_miss, outcome = diabetes) %>%
  mash() %>%
  stir() %>%
  ferment() %>%
  bottle(type = 'tibble')


sip(sft_brew, from = training, data_complete = df_cplt)

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