Calculate a median from a survey

Usage

create_analysis_median(design, group_var = NA, analysis_var, level = 0.95)

Arguments

design: design survey
group_var: dependent variable(s), variable to group by. If no dependent variable, it should be NA or empty string. If more than one variable, it should be one string with each variable separated by comma, e.g. "groupa, groupb" to group for groupa and groupb. NA is default for no grouping.
analysis_var: the independent variable, variable to summarise
level: the confidence level. 0.95 is default

Value

a data frame with the median for each group

Note

The results may differ with median(). There are lots of ways to calculate the median and the default calculation between stats::median and survey::svyquantile/srvyr::survey_median are different. Default from survey/srvyr is "school" methodology and does not exist in stats package. The default for stats is "hf7". survey/srvyr methodology is prefered as these packages are built for complex survey design.

Examples

somedata <- data.frame(
  aa = 1:10,
  bb = rep(c("a", "b"), 5),
  weights = rep(c(.5, 1.5), 5)
)
me_design <- srvyr::as_survey(somedata)
create_analysis_median(me_design, analysis_var = "aa")
#> # A tibble: 1 × 13
#>   analysis_type analysis_var analysis_var_value group_var group_var_value  stat
#>   <chr>         <chr>        <chr>              <chr>     <chr>           <dbl>
#> 1 median        aa           NA                 NA        NA                  5
#> # ℹ 7 more variables: stat_low <dbl>, stat_upp <dbl>, n <int>, n_total <dbl>,
#> #   n_w <dbl>, n_w_total <dbl>, analysis_key <chr>
create_analysis_median(me_design, group_var = "bb", analysis_var = "aa")
#> # A tibble: 2 × 13
#>   analysis_type analysis_var analysis_var_value group_var group_var_value  stat
#>   <chr>         <chr>        <chr>              <chr>     <chr>           <dbl>
#> 1 median        aa           NA                 bb        a                   5
#> 2 median        aa           NA                 bb        b                   6
#> # ℹ 7 more variables: stat_low <dbl>, stat_upp <dbl>, n <int>, n_total <dbl>,
#> #   n_w <dbl>, n_w_total <dbl>, analysis_key <chr>

me_design_w <- srvyr::as_survey(somedata, weights = weights)
create_analysis_median(me_design_w, analysis_var = "aa")
#> # A tibble: 1 × 13
#>   analysis_type analysis_var analysis_var_value group_var group_var_value  stat
#>   <chr>         <chr>        <chr>              <chr>     <chr>           <dbl>
#> 1 median        aa           NA                 NA        NA                  6
#> # ℹ 7 more variables: stat_low <dbl>, stat_upp <dbl>, n <int>, n_total <dbl>,
#> #   n_w <dbl>, n_w_total <dbl>, analysis_key <chr>
create_analysis_median(me_design_w, group_var = "bb", analysis_var = "aa")
#> # A tibble: 2 × 13
#>   analysis_type analysis_var analysis_var_value group_var group_var_value  stat
#>   <chr>         <chr>        <chr>              <chr>     <chr>           <dbl>
#> 1 median        aa           NA                 bb        a                   5
#> 2 median        aa           NA                 bb        b                   6
#> # ℹ 7 more variables: stat_low <dbl>, stat_upp <dbl>, n <int>, n_total <dbl>,
#> #   n_w <dbl>, n_w_total <dbl>, analysis_key <chr>