simpute

The goal of simpute to do Simple Imputation that works for all data types and provides a quick starting point for modelling tasks. Many other packages do more complex forms of imputation (MICE, Amelia, missForest, Hmisc, mi). Simpute is quick, simple and robust.

Installation

You can install the development version of simpute from github:

devtools::install_github("tomliptrot/simpute")

Example 1: imputation

The most basic use case is to impute all missing values in a dataframe using the impute function. This is done using the median for continous data and the mode for categorical data.

library(simpute)
colSums(is.na(airquality))

# Ozone Solar.R    Wind    Temp   Month     Day 
# 37       7       0       0       0       0 

airquality_complete = impute(airquality)
colSums(is.na(airquality_complete))

# Ozone Solar.R    Wind    Temp   Month     Day 
#  0       0       0       0       0       0 

Example 2: removing excess missing rows and columns

In some cases there is too much missing data and it might be a beter idea to completly remove either a row or column. In these cases we can use the functions remove_high_missing_row or remove_high_missing_col. The argument prop determines what proportion of data must be missing for the row/column to be removed.

library(simpute)

remove_missing_col(airquality,  prop = 0.1)

remove_missing_row(airquality,  prop = 0.1)

Example 3: piping

These functions can all be piped together using %>%

library(simpute)

airquality %>%
  remove_missing_col(prop = 0.8) %>%
  remove_missing_row(prop = 0.8)  %>%
  impute()