Calculate the an unweighted summary statistic from a survey

Calculate unweighted summaries from a survey dataset, just as on a normal data.frame with summarise. Though it is possible to use regular functions directly, because the survey package doesn't always remove rows when filtering (instead setting the weight to 0), this can sometimes give bad results. See examples for more details.

unweighted(...)

Arguments

...

variables or expressions, calculated on the unweighted data.frame behind the tbl_svy object.

Details

Uses tidy evaluation semantics and so if you want to use wrapper functions based on variable names, you must use tidy evaluation, see the examples here, documentation in nse-force, or the dplyr vignette called 'programming' for more information.

Examples

library(survey)
library(dplyr)
data(api)

dstrata <- apistrat %>%
  as_survey_design(strata = stype, weights = pw)

dstrata %>%
  summarise(api99_unw = unweighted(mean(api99)),
            n = unweighted(n()))
#> # A tibble: 1 × 2
#>   api99_unw     n
#>       <dbl> <int>
#> 1      625.   200

dstrata %>%
  group_by(stype) %>%
  summarise(api_diff_unw = unweighted(mean(api00 - api99)))
#> # A tibble: 3 × 2
#>   stype api_diff_unw
#>   <fct>        <dbl>
#> 1 E            38.6 
#> 2 H             8.46
#> 3 M            26.4 


# Some survey designs, like ones with raked weights, are not removed
# when filtered to preserve the structure. So if you don't use `unweighted()`
# your results can be wrong.
# Declare basic clustered design ----
cluster_design <- as_survey_design(
  .data = apiclus1,
  id = dnum,
  weights = pw,
  fpc = fpc
)

# Add raking weights for school type ----
pop.types <- data.frame(stype=c("E","H","M"), Freq=c(4421,755,1018))
pop.schwide <- data.frame(sch.wide=c("No","Yes"), Freq=c(1072,5122))

raked_design <- rake(
  cluster_design,
  sample.margins = list(~stype,~sch.wide),
  population.margins = list(pop.types, pop.schwide)
)

raked_design %>%
filter(cname != "Alameda") %>%
  group_by(cname) %>%
  summarize(
    direct_unw_mean = mean(api99),
    wrapped_unw_mean = unweighted(mean(api99))
  ) %>%
  filter(cname == "Alameda")
#> # A tibble: 1 × 3
#>   cname   direct_unw_mean wrapped_unw_mean
#>   <chr>             <dbl>            <dbl>
#> 1 Alameda             609              NaN

# Notice how the results are different when using `unweighted()`