surveysd (version 2.0.1)

get.selection: Get sample selection (~deltas) from drawn bootstrap replicates

Description

Reconstruct sample selection, e.g. record was drawn or not drawn (delta = 0/1) in each sampling stage from bootstrap replicates. get.selection() needs the cluster, strata and hid/pid information (if not NULL) to correctly reconstruct if a record was drawn in each sampling stage for each bootstrap replicate. Is only needed if bootstrap replicates are drawn for a survey with existing bootstrap replicates from a previous period, see parameter already.selected in function draw.bootstrap().

Usage

get.selection(
  dat,
  b.rep = attr(dat, "b.rep"),
  strata = attr(dat, "strata"),
  cluster = attr(dat, "cluster"),
  hid = attr(dat, "hid"),
  pid = attr(dat, "pid")
)

Value

Returns a list of data.tables. The length of the list equals the number of sampling stages specified. Each list entry contains a data.table with variables for sampling stage and/or hid/pid as well as length(attr(dat,"b.rep")) columns each indicating if record/cluster was drawn in the respective sampling stage for the i-th boostrap replicate.

Arguments

dat

either data.frame or data.table containing the survey data with rotating panel design. Should contain only survey data from a single time period.

b.rep

character specifying the names of the columns in dat containing bootstrap replicates.

strata

character vector specifying the name(s) of the column in dat by which the population was stratified.

cluster

character vector specifying cluster in the data.

hid

character specifying the name of the column in dat containing the household id. If NULL (the default), the household structure is not regarded. hid and pid cannot both be NULL.

pid

pid column in dat specifying the personal identifier. This identifier needs to be unique for each person throught the whole data set. hid and pid cannot both be NULL.

Examples

library(surveysd)
library(data.table)
setDTthreads(1)
set.seed(1234)
eusilc <- demo.eusilc(n = 3, prettyNames = TRUE)

## draw replicates with stratification
dat_boot <- draw.bootstrap(eusilc[year<2012], REP = 3, weights = "pWeight",
                           strata = "region", hid = "hid",
                           period = "year")

## get selection matrix for year 2011 
dat_selection <- get.selection(dat_boot[year==2011])
print(dat_selection)

## draw bootstrap replicates for year 2012
## respecting already selected units for year 2011 ~ dat_selection
## in order to mimic rotating panel design
dat_boot_2012 <- draw.bootstrap(eusilc[year==2012], REP = 3, weights = "pWeight",
                                strata = "region", hid = "hid",
                                period = "year", 
                                already.selected = dat_selection)