missEval: Evaluate imputation accuracy and consistency

In sumtxt/missEval: Functions to measure imputation accuracy and consistency

Description

Compare observed values to imputed values for a single variable

Usage

missEval(original, amputated, imputed, var, id = NULL)

Arguments

original	original data.frame
amputated	amputated data.frame
imputed	imputed data.frame
var	name of variable for evaluation
id	unique id in data.frame

Details

If a unique id is supplied, the order of all three data frames is aligned and otherwise the correct sorting is assumed. The function compares amputated values to imputed values using the Kolmogorov–Smirnov statistic and the standardized RMSE (numerical variables) or the classification error (for variable types character, factor, ordered, logical). Amputated values are observed in original but not in amputated. The RMSE is standardized with the standard deviation observed in original. While RMSE and classification error measure imputation accuracy, the Kolmogorov–Smirnov statistic measures imputation consistency.

Value

data.frame.

Examples

df <- tibble(a=1:5, b=LETTERS[1:5])
df_amp <- missMaker(df, var='a', size=3)
df_imp <- missMean(df_amp)
missEval(df, df_amp, df_imp, var='a')

sumtxt/missEval documentation built on July 12, 2020, 12:07 a.m.