missMethods (version 0.2.0)

impute_mean: Mean imputation

Description

Impute an observed mean for the missing values

Usage

impute_mean(ds, type = "columnwise")

Arguments

ds

A data frame or matrix with missing values.

type

One of: "columnwise", "rowwise", "total", "Two-Way" or "Winer" (see details).

Value

An object of the same class as ds with imputed missing values.

A Note for tibble users

If you use tibbles and an error like ‘Lossy cast from `value` double to integer’ occurs, you will first need to convert all integer columns with missing values to double. Another solution is to convert the tibble with as.data.frame() to a data frame. The data frame will automatically convert integer columns to double columns, if needed.

Details

For every missing value the mean of some observed values is imputed. The observed values to be used are specified via type. For example, type = "columnwise" (the default) imputes the mean of the observed values in a column for all missing values in the column. This is normally meant, if someone speaks of "imputing the mean" or "mean imputation".

Other options for type are: "rowwise", "total", "Winer" and "Two-way". The option "rowwise" imputes all missing values in a row with the mean of the observed values in the same row. "total" will impute every missing value with the mean of all observed values in ds. "Winer" imputes the mean of the rowwise and columnwise mean. Beland et al. (2016) called this method "Winer" and they attributed the method to Winer (1971). "Two-way" imputes the sum of rowwise and columnwise mean minus the total mean. This method was suggested by D.B Rubin to Bernaards & Sijtsma, K. (2000).

References

Beland, S., Pichette, F., & Jolani, S. (2016). Impact on Cronbach's \(\alpha\) of simple treatment methods for missing data. The Quantitative Methods for Psychology, 12(1), 57-73.

Bernaards, C. A., & Sijtsma, K. (2000). Influence of imputation and EM methods on factor analysis when item nonresponse in questionnaire data is nonignorable. Multivariate Behavioral Research, 35(3), 321-364.

Winer, B. J. (1971). Statistical principles in experimental design (2ed ed.) New York: McGraw-Hill

See Also

apply_imputation the workhorse for this function.

Other location parameter imputation functions: impute_median(), impute_mode()

Examples

# NOT RUN {
ds <- data.frame(X = 1:20, Y = 101:120)
ds_mis <- delete_MCAR(ds, 0.2)
ds_imp <- impute_mean(ds_mis)
# completely observed columns can be of any type:
ds_mis_char <- cbind(ds_mis, letters[1:20])
ds_imp_char <- impute_mean(ds_mis_char)
# }