From b37aa3251a75d79bbc37e3d45ddb336034054147 Mon Sep 17 00:00:00 2001
From: louis-del <louis.delamarre.pro@gmail.com>
Date: Tue, 22 Aug 2023 23:25:09 +0200
Subject: [PATCH] 2023-08-22: Added a function to add a column to table1
 displaying the effect sizes of the differences between groups. For numeric
 variables, if equal variances, Cohen's d is applied, if variances are
 unequal, Glass' delta is applied. For categorical variables, Cramer's V is
 used and the degrees of freedom are displayed in the output to facilitate
 interpretation.

---
 R/table1_effect-size.R | 56 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 56 insertions(+)
 create mode 100644 R/table1_effect-size.R

diff --git a/R/table1_effect-size.R b/R/table1_effect-size.R
new file mode 100644
index 0000000..c597cb1
--- /dev/null
+++ b/R/table1_effect-size.R
@@ -0,0 +1,56 @@
+
+
+##create a function to add effect size to the table1
+
+# Load necessary libraries
+library(effectsize)
+library(vcd)
+
+# Create a function to add effect size to table1 package results
+effect_size_table1 <- function(x, ...) {
+  # Construct vectors of data y, and groups (strata) g
+  y <- unlist(x)
+  g <- factor(rep(1:length(x), times=sapply(x, length)))
+  
+  if (is.numeric(y)) {
+    # For numeric variables, check if variances differ between the groups
+    levene_test <- car::leveneTest(y, g)
+    p_val_levene <- levene_test[["Pr(>F)"]][[1]]
+    var_equal <- p_val_levene > 0.05
+    
+    if (var_equal) {
+      # Use Cohen's d if variances do not differ between the groups
+      es <- cohens_d(y ~ g)$Cohens_d
+      test_name <- "Cohen's d"
+      interp <- ifelse(abs(es) < 0.2, "Neglig.", 
+                       ifelse(abs(es) < 0.5, "Small", 
+                              ifelse(abs(es) < 0.8, "Medium", "Large")))
+    } else {
+      # Use Glass's delta if variances differ between the groups
+      es <- glass_delta(y ~ g)$Glass_delta
+      test_name <- "Glass' Delta"
+      interp <- ifelse(abs(es) < 0.2, "Neglig.", 
+                       ifelse(abs(es) < 0.5, "Small", 
+                              ifelse(abs(es) < 0.8, "Medium", "Large")))
+    }
+  } else {
+    # For categorical variables, use Cramer's V
+    es <- cramers_v(table(y, g))$Cramers_v
+    dF <- min(dim(table(y, g))[1]-1, dim(table(y, g))[2]-1)
+    test_name <- paste0("Cramer's V - dF (", dF, ")") 
+    interp <- ifelse(abs(es) < 0.1, "Neglig.", 
+                     ifelse(abs(es) < 0.3, "Small", 
+                            ifelse(abs(es) < 0.5, "Medium", "Large")))
+  }
+  
+
+  # Format the effect size and include the name of the test applied
+  result <- paste0(format(round(es, 3), nsmall = 3), " (", test_name, ") ", interp)
+  # The initial empty string places the output on the line below the variable label.
+  c("", result)
+}
+
+
+#how to use it : 
+#my_table1 <- table1::table1(~  predictor1 + predictor2 + predictor3 | Outcome, data = data, overall=F, extra.col=list(`P-value`=pvalue, `Effect size`=effect_size_table1))
+