Merge pull request #845 from vincentarelbundock/issue840

issue #840 blocks, clusters, and weights in datasummary_balance()

DESCRIPTION

@@ -9,7 +9,7 @@ Description: Create beautiful and customizable tables to summarize several
     RTF, JPG, or PNG. Tables can easily be embedded in 'Rmarkdown' or 'knitr'
     dynamic documents. Details can be found in Arel-Bundock (2022)
     <doi:10.18637/jss.v103.i01>.
-Version: 2.2.0.6
+Version: 2.2.0.7
 Remotes: vincentarelbundock/tinytable
 Authors@R: c(person("Vincent", "Arel-Bundock", 
                    email = "[email protected]", 
@@ -111,6 +111,7 @@ Suggests:
     parallel,
     pscl,
     psych,
+    randomizr,
     remotes,
     rmarkdown,
     rstanarm,

NEWS.md

@@ -8,6 +8,7 @@ Bugs:
 * Bad horizontal rule placement with `add_rows`. Thanks to @pyoungblood for Issue #813.
 * `shape` creates bad columns when model names include spaces. Thanks to @daSilva5 for report #816.
 * Center align GOF in dot-aligned columns. Thanks to @svraka for report #827.
+* Weights, blocks, and clusters ignored by `datasummary_balance()`. Thanks to @kleuveld for report #840.
 
 New features:
 

R/datasummary_balance.R

@@ -393,8 +393,9 @@ sanitize_datasummary_balance_data <- function(formula, data) {
   lhs <- stats::update(formula, ".~1")
   lhs <- all.vars(lhs)
   lhs <- setdiff(lhs, ".")
+  bonus <- c("clusters", "weights", "blocks")
   if (length(lhs) > 0) {
-    cols <- intersect(c(lhs, rhs), colnames(data))
+    cols <- intersect(c(lhs, rhs, bonus), colnames(data))
     if (length(cols) > 1) {
       data <- data[, cols, drop = FALSE]
     }

inst/tinytest/test-datasummary_balance.R

@@ -327,3 +327,32 @@ expect_snapshot_print(tab, "datasummary_balance-issue711")
 tab <- datasummary_balance(mpg + hp ~ am, data = mtcars, output = "dataframe")
 expect_equal(nrow(tab), 2)
 
+
+# Issue #840
+requiet("randomizr")
+requiet("estimatr")
+requiet("tibble")
+
+dat <- tibble::tibble(
+  x = rnorm(n = 1000),
+  clusters = randomizr::simple_ra(N = 1000, num_arms = 50),
+  Z = randomizr::cluster_ra(clusters = clusters),
+  y = x + Z * 5,
+  weights = ifelse(y > 5, 0, 1)) # making sure the weights actually make a difference
+dat <- data.frame(dat)
+tab <- datasummary_balance(y ~ Z, data = dat, fmt = 4, output = "data.frame")
+
+wm <- weighted.mean(dat$y[dat$Z == 1], dat$weights[dat$Z == 1])
+wm <- sprintf("%.4f", wm)
+expect_equivalent(wm, tab[, 4])
+wm <- weighted.mean(dat$y[dat$Z == 0], dat$weights[dat$Z == 0])
+wm <- sprintf("%.4f", wm)
+expect_equivalent(wm, tab[, 2])
+
+# and these should be the differences
+dinm <- difference_in_means(y ~ Z,
+  clusters = clusters,
+  weights = weights,
+  data = dat)
+dinm <- sprintf("%.4f", coef(dinm))
+expect_equivalent(dinm, tab[, 6])