A one-stage design: sampling weights, optional strata, optional primary sampling units
(clusters) and an optional finite-population correction. Everything here is native; the
variance of the estimators built on it is the Taylor linearisation of survey::svyrecvar.
Usage
morie_survey_design(
data,
weights_col,
strata_col = NULL,
cluster_col = NULL,
fpc_col = NULL,
nest = FALSE
)Arguments
- data
data.frame.
- weights_col
Column name of the sampling weights (inverse inclusion probabilities).
- strata_col
Optional strata column.
- cluster_col
Optional PSU/cluster column.
- fpc_col
Optional finite-population-correction column: the population size of the stratum (number of PSUs when clustered), or the sampling fraction when every value is <= 1.
- nest
If TRUE, cluster IDs are only unique within a stratum; if FALSE (the default) a cluster ID that appears in two strata is an error, as in
survey::svydesign.
Value
An object of class "morie_survey_design": a list with the data, the
weights, and per-row stratum, PSU, PSU count of the stratum and population size.
Examples
set.seed(1)
df <- data.frame(y = rnorm(40), w = runif(40, 0.5, 2),
s = rep(c("a", "b"), 20), psu = rep(1:10, each = 4))
d <- morie_survey_design(df, "w", strata_col = "s", cluster_col = "psu", nest = TRUE)
class(d)
#> [1] "morie_survey_design" "morie_survey_design_fallback"
morie_survey_mean(d, "y")
#> $mean
#> [1] 0.07347258
#>
#> $se
#> [1] 0.1481628
#>