NAReplace: NA Replace

In mattmills49/modeler: Exploratory Data Analysis and Modeling Helper Functions

Description

This function takes in an x and y vector and attempts to replace missing values. Replacement can be done with the mean or median value of the non-missing values or using a regression replace technique. The regression technique first finds the predicted response value of y for the non missing elements of x. The average value of y for the missing values is then used to find the x-value that most closely aligns with this average value of y for the missing values.

Usage

NAReplace(x, y, replace = "linear")

Arguments

x	independent variable, must be a vector. If x is a factor it will be converted to a character vector
y	dependent variable, must be a vector and can't contain missing values
replace	The type of NA replacement method. Options include mean, median, linear, sample, and smooth.

Value

A list containing x_orig The original x vector x_replace The new vector with missing values filled in na_indicator A dummy variable (0 or 1) with that tells you the indices of missing values replacement_val The value that the means were replaced with

Examples

y <- c(sample(c(0, 1), 25, replace = T, prob = c(.25, .75)), sample(c(0, 1), 25, replace = T, prob = c(.4, .6)), sample(c(0, 1), 25, replace = T, prob = c(.6, .4)), sample(c(0, 1), 25, replace = T, prob = c(.75, .25)))
set.seed(924516)
x_real <- (seq(1, 10, length.out = 100))^2 + runif(100, 0, 5)
x <- x_real
x[1:50][runif(50) < .33] <- NA
nafix <- NAReplace(x, y, replace = "mean")
nafix$replacement_val
# [1] 45.56847
nafix <- NAReplace(x, y, replace = "linear")
nafix$replacement_val
# [1] 11.80703
x_char <- c(sample(letters[1:10], 50, replace = T), sample(letters[1:5], 50, replace = T))
x_char[50:100][runif(50) < .25] <- NA
nafix <- NAReplace(x_char, y, replace = "median")
nafix$replacement_val
# [1] "c"
nafix <- NAReplace(x_char, y, replace = "linear")
nafix$replacement_val
# [1] "d"

mattmills49/modeler documentation built on May 21, 2019, 1:25 p.m.