bootVar {laeken}

Bootstrap variance and confidence intervals of indicators on social exclusion and poverty

Description

Compute variance and confidence interval estimates of indicators on social exclusion and poverty based on bootstrap resampling.

Usage

bootVar(
  inc,
  weights = NULL,
  years = NULL,
  breakdown = NULL,
  design = NULL,
  cluster = NULL,
  data = NULL,
  indicator,
  R = 100,
  bootType = c("calibrate", "naive"),
  X,
  totals = NULL,
  ciType = c("perc", "norm", "basic"),
  alpha = 0.05,
  seed = NULL,
  na.rm = FALSE,
  gender = NULL,
  method = NULL,
  ...
)

Arguments

inc

either a numeric vector giving the equivalized disposable income, or (if data is not NULL) a character string, an integer or a logical vector specifying the corresponding column of data.

weights

optional; either a numeric vector giving the personal sample weights, or (if data is not NULL) a character string, an integer or a logical vector specifying the corresponding column of data.

years

optional; either a numeric vector giving the different years of the survey, or (if data is not NULL) a character string, an integer or a logical vector specifying the corresponding column of data. If supplied, values are computed for each year.

breakdown

optional; either a numeric vector giving different domains, or (if data is not NULL) a character string, an integer or a logical vector specifying the corresponding column of data. If supplied, the values for each domain are computed in addition to the overall value.

design

optional; either an integer vector or factor giving different strata for stratified sampling designs, or (if data is not NULL) a character string, an integer or a logical vector specifying the corresponding column of data. If supplied, this is used as strata argument in the call to boot.

cluster

optional; either an integer vector or factor giving different clusters for cluster sampling designs, or (if data is not NULL) a character string, an integer or a logical vector specifying the corresponding column of data.

data

an optional data.frame.

indicator

an object inheriting from the class "indicator" that contains the point estimates of the indicator (see arpr, qsr, rmpg or gini).

R

a numeric value giving the number of bootstrap replicates.

bootType

a character string specifying the type of bootstap to be performed. Possible values are "calibrate" (for calibration of the sample weights of the resampled observations in every iteration) and "naive" (for a naive bootstrap without calibration of the sample weights).

X

if bootType is "calibrate", a matrix of calibration variables.

totals

numeric; if bootType is "calibrate", this gives the population totals. If years is NULL, a vector should be supplied, otherwise a matrix in which each row contains the population totals of the respective year. If this is NULL (the default), the population totals are computed from the sample weights using the Horvitz-Thompson estimator.

ciType

a character string specifying the type of confidence interval(s) to be computed. Possible values are "perc", "norm" and "basic" (see boot.ci).

alpha

a numeric value giving the significance level to be used for computing the confidence interval(s) (i.e., the confidence level is 1 - alpha), or NULL.

seed

optional; an integer value to be used as the seed of the random number generator, or an integer vector containing the state of the random number generator to be restored.

na.rm

a logical indicating whether missing values should be removed.

gender

either a numeric vector giving the gender, or (if data is not NULL) a character string, an integer or a logical vector specifying the corresponding column of data.

method

a character string specifying the method to be used (only for gpg). Possible values are "mean" for the mean, and "median" for the median. If weights are provided, the weighted mean or weighted median is estimated.

...

if bootType is "calibrate", additional arguments to be passed to calibWeights.

Value

An object of the same class as indicator is returned. See arpr, qsr, rmpg or gini for details on the components.

Note

This function gives reasonable variance estimates for basic sample designs such as simple random sampling or stratified simple random sampling.

Author(s)

Andreas Alfons

References

A. Alfons and M. Templ (2013) Estimation of Social Exclusion Indicators from Complex Surveys: The R Package laeken. Journal of Statistical Software, 54(15), 1–25. doi:10.18637/jss.v054.i15

See Also

variance, calibWeights, arpr, qsr, rmpg, gini

Examples

data(eusilc)
a <- arpr("eqIncome", weights = "rb050", data = eusilc)

## naive bootstrap
bootVar("eqIncome", weights = "rb050", design = "db040",
    data = eusilc, indicator = a, R = 50,
    bootType = "naive", seed = 123)

## bootstrap with calibration
bootVar("eqIncome", weights = "rb050", design = "db040",
    data = eusilc, indicator = a, R = 50,
    X = calibVars(eusilc$db040), seed = 123)