na_predict: Replace missing values using multivariate statistical model

In twolodzko/misster: Simple tools for dealing with missing values

Description

Replace missing values using multivariate statistical model

Usage

na_predict(data, formula, learnFun, predictFun = predict, ...)

na_lm(data, formula, ...)

na_glm(data, formula, family = gaussian, ...)

na_rpart(data, formula, ...)

na_rf(data, formula, ...)

na_nb(data, formula, ...)

na_knn(data, formula, ...)

Arguments

data	a data.frame.
formula	an object of class "formula": a symbolic description of the model to be fitted.
learnFun	learning function in form learnFun(formula, data, ...).
predictFun	function used for making predictions in form predictFun(object, newdata).
...	further arguments passed to learnFun.
family	in na_glm, this is the family argument from the glm method.

Details

Multiple convenience wrappers allow user to use: linear regression (na_lm), generalized linear models (na_glm), recursive partitioning and regression trees (na_rpart), random forests (na_rf) and additionally, for categorical data: naive Bayes (na_nb) and k-nearest neighbour classifiers (na_knn). Both na_rpart and na_rf can be used for predicting continuous and categorical variables.

See Also

lm, glm, rpart, randomForest, naiveBayes, knn

Examples

set.seed(123)

dat <- mtcars
dat$disp[sample.int(nrow(dat), 10)] <- NA
dat$gear[sample.int(nrow(dat), 10)] <- NA
dat$gear <- as.factor(dat$gear)

na_predict(dat, disp ~ mpg + drat, learnFun = glm, predictFun = function(object, newdata) {
           predict(object, newdata= newdata, type = "response") })
na_predict(dat, gear ~ mpg + drat, learnFun = e1071::naiveBayes)

# continuous variables
na_lm(dat, disp ~ mpg + drat)
na_glm(dat, disp ~ mpg + drat)
na_rpart(dat, disp ~ mpg + drat)
na_rf(dat, disp ~ mpg + drat)

# categorical variables
na_nb(dat, gear ~ mpg + drat)
na_knn(dat, gear ~ mpg + drat)
na_rpart(dat, factor(gear) ~ mpg + drat)
na_rf(dat, factor(gear) ~ mpg + drat)

twolodzko/misster documentation built on May 24, 2019, 2:54 p.m.