Weighted effect sizes

[mattansb]

Weighting is already implemented in standardize.default() and standardize_parameters() (and eta_squared() via the fit models).

Would be nice to have this also (with weights = arg) in:

• [ ] cohens_d() etc
• [ ] cramers_v() etc
• [ ] Rank effect sizes

Perhaps we can ~steal~ borrow some code from sjstats::weighted_*...

[mattansb]

For all the Xsq stuff, this seems easy - essentially, all suppressWarnings(stats::chisq.test(x, y, ...)) should be replaced with:

chisq.test2 <- function(x, y, w = NULL, ...) {
  if (!is.null(w)) {
    x <- stats::xtabs(w ~ x + y, data = mtcars)
    y <- NULL
  }

  suppressWarnings(stats::chisq.test(x, y, ...))
}

chisq.test2(mtcars$am, mtcars$cyl)
#> 
#>  Pearson's Chi-squared test
#> 
#> data:  x and y
#> X-squared = 8.7407, df = 2, p-value = 0.01265

chisq.test2(mtcars$am, mtcars$cyl, w = mtcars$gear)
#> 
#>  Pearson's Chi-squared test
#> 
#> data:  x
#> X-squared = 26.383, df = 2, p-value = 1.866e-06

^{Created on 2021-11-30 by the reprex package (v2.0.1)}