Quick and dirty imputation of missing values.

Description

Imputes missing values by fitting an additive or multiplicative main effects model to the non-missing data, and then randomly drawing values for the missing values, based on predictions from the model.

Usage

impute(object, mult = NULL, max = NULL)

## S4 method for signature 'CountsWithSubtotals'
impute(object, mult = NULL, max = NULL)

## S4 method for signature 'DemographicArray'
impute(object, mult = NULL, max = NULL)

Arguments

Details

impute performs a single imputation based on a simple, and not necessarily appropriate, model. impute is useful for generating starting values for iterative calculations, for tidying data with a small proportion of missing values, or for rough approximations. For more sophisticated approaches to the imputation of demographic data see package demest.

If mult is TRUE, meaning that the multiplicative model is used, all non-missing values must be non-negative (0s are omitted from the model.) If the mult is TRUE all imputed values are guaranteed to be non-negative.

If all non-missing values are non-negative integers, imputed values are drawn from a Poisson distribution. Other they are drawn from a normal distribution.

The max argument is useful when object represents draws from a binomial distribution. Setting max to the number of trials for each cell ensures that the imputed number of successes does not exceed the number of trials.

Value

An object with the same class as object, with the missing values filled in.

References

Little, RJA and Rubin, DB. 2002. Statistical Analysis with Missing Data, Wiley.

See Also

DemographicArray, perturb, extrapolate

Examples

library(demdata)
popn <- Counts(VAPopn)
popn[sample(length(popn), size = 5)] <- NA
popn
impute(popn)
impute(popn, mult = TRUE)
impute(popn, max = 500)