Impute Missing Values

There are two imputation strategies:

Imputing with the median is typically performed for numeric vectors, however methods also exist for:

• the factor method imputes to the most common factor level.

• the character method imputes to the most common string

Imputing missing values from an estimated distribution is performed by fitting via Maximum Likelihood and estimating and parameters, followed by random draws from the appropriate random generator. Currently only supports Gamma, Gaussian, and Beta distributions.

Usage

impute_median(x)

impute_missing(x, distr = c("norm", "gamma", "beta"), seed = 123)

Arguments

x

A data.frame, matrix, or numeric vector. Factors or character strings are also possible but see caveat above.

distr

character(1). The name of the desired distribution.

seed

integer(1). Set a seed for reproducibility.

Value

An object of the same dimension and class as x with any NA (missing) values replaced.

See also

median()

Other impute: impute_outliers(), impute_predictors()

Author

Stu Field

Examples

# data frame method
df <- select(mtcars, mpg, cyl, disp) |> head()
df[cbind(2:3, 2:3)] <- NA_real_
df
#>                    mpg cyl disp
#> Mazda RX4         21.0   6  160
#> Mazda RX4 Wag     21.0  NA  160
#> Datsun 710        22.8   4   NA
#> Hornet 4 Drive    21.4   6  258
#> Hornet Sportabout 18.7   8  360
#> Valiant           18.1   6  225

impute_median(df)
#>                    mpg cyl disp
#> Mazda RX4         21.0   6  160
#> Mazda RX4 Wag     21.0   6  160
#> Datsun 710        22.8   4  225
#> Hornet 4 Drive    21.4   6  258
#> Hornet Sportabout 18.7   8  360
#> Valiant           18.1   6  225
x <- rnorm(10, mean = 10, sd = 5)
x[c(2, 7)] <- NA_real_
y <- rbeta(10, shape1 = 2, shape2 = 3)
y[c(2, 7)] <- NA_real_

data.frame(
  orig_x    = x,
  imputed_x = impute_missing(x), # gaussian
  orig_y    = y,
  imputed_y = impute_missing(y, "beta") # beta
)
#>       orig_x imputed_x     orig_y  imputed_y
#> 1   7.105577  7.105577 0.33386926 0.33386926
#> 2         NA  9.971782         NA 0.18073603
#> 3  10.664961 10.664961 0.41652882 0.41652882
#> 4  11.882497 11.882497 0.16576899 0.16576899
#> 5  15.693538 15.693538 0.25369523 0.25369523
#> 6  16.206315 16.206315 0.04883205 0.04883205
#> 7         NA 11.156292         NA 0.23837690
#> 8   7.853100  7.853100 0.31795159 0.31795159
#> 9  16.802307 16.802307 0.60805309 0.60805309
#> 10  9.645713  9.645713 0.13826012 0.13826012