eff_n {svyweight}

Effective Sample Size and Weighting Efficiency

Description

Computes Kish's effective sample size or weighting efficiency for a survey.design object.

Usage

eff_n(design)

weight_eff(design)

Arguments

design

An svydesign object, presumably with design or post-stratification weights.

Details

Kish's effective sample size is a frequently-used, general metric to indicate how much uncertainty and error increase due to weighting. Effective sample size is calculated as sum(weights(design))^2 / sum(weights(design)^2). Weighting efficiency is eff_n(design) / sum(weights(design)).

While weighting efficency and effective sample size are frequently use, they are less valid than the standard errors produced by survey::svymean() and related functions from the survey package. In particular, they ignore clustering and stratification in sample designs, and covariance between weighting variables and outcome variables. As such, these metrics should be used with caution

Value

A numeric value, indicating effective sample size (for eff_n()) or weighting efficiency (for weight_eff())

References

Kish, Leslie. 1965. Survey Sampling New York: Wiley.

Examples

gles17_weighted <- rakesvy(design = gles17, 
    gender ~ c("Male" = .495, "Female" = .505),
    eastwest ~ c("East Germany" = .195, "West Germany" = .805)
)

eff_n(gles17_weighted)
weight_eff(gles17_weighted)