Error estimation

Prerequisites

In order to use a dataset with calc.stError(), several weight columns have to be present. Each weight column corresponds to a bootstrap sample. In the following examples, we will use the data from demo.eusilc() and attach the bootstrap weights using draw.bootstrap() and recalib(). Please refer to the documentation of those functions for more detail.

library(surveysd)

set.seed(1234)
eusilc <- demo.eusilc(prettyNames = TRUE)
dat_boot <- draw.bootstrap(eusilc, REP = 10, hid = "hid", weights = "pWeight",
                           strata = "region", period = "year")
dat_boot_calib <- recalib(dat_boot, conP.var = "gender", conH.var = "region",
                          epsP = 1e-2, epsH = 2.5e-2, verbose = FALSE)
dat_boot_calib[, onePerson := nrow(.SD) == 1, by = .(year, hid)]

## print part of the dataset
dat_boot_calib[1:5, .(year, povertyRisk, eqIncome, onePerson, pWeight, w1, w2, w3, w4, w5)]

Estimator functions

The parameters fun and var in calc.stError() define the estimator to be used in the error analysis. There are two built-in estimator functions weightedSum() and weightedRatio() which can be used as follows.

povertyRate <- calc.stError(dat_boot_calib, var = "povertyRisk", fun = weightedRatio)
totalIncome <- calc.stError(dat_boot_calib, var = "eqIncome", fun = weightedSum)

Those functions calculate the ratio of persons at risk of poverty (in percent) and the total income. By default, the results are calculated separately for each reference period.

povertyRate$Estimates

totalIncome$Estimates

Columns that use the val_ prefix denote the point estimate belonging to the “main weight” of the dataset, which is pWeight in case of the dataset used here.

Columns with the stE_ prefix denote standard errors calculated with bootstrap replicates. The replicates result in using w1, w2, …, w10 instead of pWeight when applying the estimator.

n denotes the number of observations for the year and N denotes the total weight of those persons.

## define custom estimator
myWeightedSum <- function(x, w) {
  sum(x*w)
}

## check if results are equal to the one using `suveysd::weightedSum()`
totalIncome2 <- calc.stError(dat_boot_calib, var = "eqIncome", fun = myWeightedSum)
all.equal(totalIncome$Estimates, totalIncome2$Estimates)

## [1] TRUE

multipleRates <- calc.stError(dat_boot_calib, var = c("povertyRisk", "onePerson"), fun = weightedRatio)
multipleRates$Estimates

Grouping

The groups argument can be used to calculate estimators for different subsets of the data. This argument can take the grouping variable as a string that refers to a column name (usually a factor) in dat. If set, all estimators are not only split by the reference period but also by the grouping variable. For simplicity, only one reference period of the above data is used.

dat2 <- subset(dat_boot_calib, year == 2010)
for (att  in c("period", "weights", "b.rep"))
  attr(dat2, att) <- attr(dat_boot_calib, att)

To calculate the ratio of persons at risk of poverty for each federal state of Austria, group = "region" can be used.

povertyRates <- calc.stError(dat2, var = "povertyRisk", fun = weightedRatio, group = "region")
povertyRates$Estimates

The last row with region = NA denotes the aggregate over all regions. Note that the columns N and n now show the weighted and unweighted number of persons in each region.

povertyRates <- calc.stError(dat2, var = "povertyRisk", fun = weightedRatio, 
                             group = c("gender", "region"))
povertyRates$Estimates

povertyRates <- calc.stError(dat2, var = "povertyRisk", fun = weightedRatio, 
                             group = list(c("gender", "region")))
povertyRates$Estimates

povertyRates <- calc.stError(dat2, var = "povertyRisk", fun = weightedRatio, 
                             group = list("gender", "region", c("gender", "region")))
povertyRates$Estimates