SciencesPo (version 1.3.8)

winsorize: Winsorized Mean

Description

Compute the winsorized mean, which consists of recoding the top k values in a vector.

Usage

winsorize(x, k = 1, na.rm = TRUE)

Arguments

x

The vector to be winsorized

k

An integer for the quantity of outlier elements that to be replaced in the calculation process

na.rm

a logical value for na.rm, default is na.rm=TRUE.

Value

• An object of the same type as x

encoding

UTF-8

Details

Winsorizing a vector will produce different results than trimming it. While by trimming a vector causes extreme values to be discarded, by winsorizing it in the other hand, causes extreme values to be replaced by certain percentiles.

References

Dixon, W. J., and Yuen, K. K. (1999) Trimming and winsorization: A review. The American Statistician, 53(3), 267--269.

Dixon, W. J., and Yuen, K. K. (1960) Simplified Estimation from Censored Normal Samples, The Annals of Mathematical Statistics, 31, 385--391.

Wilcox, R. R. (2012) Introduction to robust estimation and hypothesis testing. Academic Press, 30-32. Statistics Canada (2010) Survey Methods and Practices.

Examples

set.seed(51)  # for reproducibility
x <- rnorm(50)
## introduce outliers
x[1] <- x[1] * 10
# Compare to mean:
 mean(x)
 winsorize(x)