Perform prediction-correction of the dependent variable

Source: R/plot_vpc_exactbins.R

df_pcdv is a helper function to perform prediction-correction of observed or simulated depedent variables.

df_pcdv(
  data,
  bin_var = "NTIME",
  strat_vars = NULL,
  output_vars = c(PRED = "PRED", IPRED = "IPRED", DV = "DV"),
  lower_bound = 0
)

Arguments

data

Input dataset

bin_var

Exact binning variable. Default is "NTIME".

strat_vars

Stratifying variables. Default is "CMT".

output_vars

Names of model outputs from model. Must be named character vector. Defaults are "PRED", "IPRED", and "DV". Currently both "IPRED" and "DV" are required in model@capture.

lower_bound

Lower bound of the dependent variable for prediction correction. Default is 0.

Value

A data.frame containing one row per unique combination of bin_var and strat_vars and new variable PCDV containing prediction-corrected observations.

Examples

model <- model_mread_load(model = "model")
#> Loading model from cache.
data <- df_addpred(data_sad, model)
simout <- df_pcdv(data, output_vars = c(DV = "ODV", PRED = "PRED"))