Fix paired check

NAMESPACE

@@ -15,3 +15,4 @@ export(median_diff)
 importFrom(ggplot2,.pt)
 importFrom(magrittr,"%>%")
 importFrom(rlang,.data)
+importFrom(dplyr, n)

R/001_api.R

@@ -232,12 +232,19 @@ load <- function(
 
   ## Check to ensure control & treatment groups have the same sample size if is_paired is TRUE
   if (is_paired) {
-    if (length(unique(Ns$n)) > 1) {
-      cli::cli_abort(c("{.field data} is paired, as indicated by {.field paired} but size of control and treatment groups are not equal.",
+    for(i in idx) {
+      unlisted_pair <- unlist(i)
+      filter_df <- data %>%
+        dplyr::filter(!!enquo_x %in% unlisted_pair)
+      check_ns <- filter_df %>%
+        dplyr::count(!!enquo_x)
+      if (length(unique(check_ns$n)) > 1) {
+        cli::cli_abort(c("{.field data} is paired, as indicated by {.field paired} but size of control and treatment groups are not equal.",
         "x" = "Ensure that the size of control and treatment groups are the same for paired comparisons."
       ))
     }
   }
+}
 
   # Extending ylim for plotting
   ylim[1] <- ylim[1] - (ylim[2] - ylim[1]) / 25

tests/testthat/test_001_api.R

@@ -1,3 +1,21 @@
+describe("Given non-valid params", {
+  # since this is more of an integration test, this unit test logic is repeated in test_001_utils.R
+  np_dataset <- generate_non_proportional_dataset()
+  p_dataset <- generate_proportional_dataset()
+
+  # for paired data test
+  np_dataset_trunc <- np_dataset[-2, ]
+
+  test_that("it should detect invalid x value", {
+    expect_error(
+      dabestr::load(np_dataset,
+        x = Grou, y = Measurement,
+        idx = c("Control 1", "Test 1")
+      ),
+      regexp = "Column x is not in data"
+    )
+  })
+
 describe("Testing load function", {
   describe("Given valid params", {
     describe("Given non proportional dataset", {
@@ -36,20 +54,7 @@ describe("Testing load function", {
     })
   })
 
-  describe("Given non-valid params", {
-    # since this is more of an integration test, this unit test logic is repeated in test_001_utils.R
-    np_dataset <- generate_non_proportional_dataset()
-    p_dataset <- generate_proportional_dataset()
 
-    test_that("it should detect invalid x value", {
-      expect_error(
-        dabestr::load(np_dataset,
-          x = Grou, y = Measurement,
-          idx = c("Control 1", "Test 1")
-        ),
-        regexp = "Column x is not in data"
-      )
-    })
 
     test_that("it should detect invalid y value", {
       expect_error(
@@ -147,7 +152,7 @@ describe("Testing load function", {
 
       test_that("it should detect invalid dataset size", {
         expect_error(
-          dabestr::load(np_dataset[-2, ],
+          dabestr::load(np_dataset_trunc,
             x = Group, y = Measurement,
             idx = c("Control 1", "Test 1", "Test 2"), paired = "sequential",
             id_col = ID

tests/testthat/test_001_utils.R

@@ -3,6 +3,9 @@ describe("Testing validate_load_params function", {
   np_dataset <- generate_non_proportional_dataset()
   p_dataset <- generate_proportional_dataset()
 
+  # for paired data test
+  np_dataset_trunc <- np_dataset[-2, ]  
+
   test_that("it should detect invalid x value", {
     expect_error(
       dabestr::load(np_dataset,
@@ -109,7 +112,7 @@ describe("Testing validate_load_params function", {
 
     test_that("it should detect invalid dataset size", {
       expect_error(
-        dabestr::load(np_dataset[-2, ],
+        dabestr::load(np_dataset_trunc,
           x = Group, y = Measurement,
           idx = c("Control 1", "Test 1", "Test 2"), paired = "sequential",
           id_col = ID

vignettes/.gitignore

@@ -1,2 +1,4 @@
 *.html
 *.R
+
+/.quarto/

vignettes/plot_aesthetics.Rmd

@@ -14,6 +14,7 @@ knitr::opts_chunk$set(
   collapse = TRUE,
   comment = "#>"
 )
+library(dplyr)
 ```
 
 Controlling plot aesthetics is straightforward in `dabestr`. A key feature of `dabestr` is the ability for users to freely customize various components of a `dabestr` estimation plot, allowing for the creation of an optimal-looking plot.

vignettes/tutorial_basics.Rmd

@@ -21,6 +21,7 @@ The dataset is first processed into the dabestr format using the `load()` functi
 
 ```{r setup, warning = FALSE, message = FALSE}
 library(dabestr)
+library(dplyr)
 ```
 
 ## Create dataset for demo

vignettes/tutorial_deltadelta.Rmd

@@ -16,6 +16,7 @@ knitr::opts_chunk$set(
 
 ```{r, include = FALSE, warning = FALSE, message = FALSE}
 library(dabestr)
+library(dplyr)
 ```
 
 This vignette documents how `dabestr` is able to compute the calculation of delta-delta, an experimental function that allows the comparison between two bootstrapped effect sizes computed from two independent categorical variables.

vignettes/tutorial_minimeta.Rmd

@@ -33,6 +33,7 @@ For more information on meta-analysis, please refer to [Chapter 10](https://trai
 
 ```{r setup, warning = FALSE, message = FALSE}
 library(dabestr)
+library(dplyr)
 ```
 
 ## Create dataset for demo

vignettes/tutorial_proportion_plots.Rmd

@@ -24,6 +24,7 @@ This vignette documents how the `dabestr` package is able to generate proportion
 
 ```{r setup, warning = FALSE, message = FALSE}
 library(dabestr)
+library(dplyr)
 ```
 
 ## Create dataset for demo

vignettes/tutorial_repeated_measures.Rmd

@@ -25,6 +25,7 @@ To use these features, simply declare the `paired` argument as either "sequentia
 
 ```{r setup, warning = FALSE, message = FALSE}
 library(dabestr)
+library(dplyr)
 ```
 
 ## Create dataset for demo
@@ -221,3 +222,52 @@ dabest_plot(multi_baseline_repeated_measures.mean_diff,
   raw_marker_size = 0.5, raw_marker_alpha = 0.3
 )
 ```
+
+## Paired groups with different sample sizes across comparisons
+
+Repeated-measures plots also work when different comparison pairs have different sample sizes, as long as each pair is internally balanced.
+
+```{r}
+set.seed(12345)
+N1 <- 20
+N2 <- 10
+
+df_unequal_pairs <- tibble::tibble(
+  Group = rep(c("Control 1", "Test 1", "Control 2", "Test 2"), c(N1, N1, N2, N2)),
+  Measurement = c(
+    rnorm(N1, mean = 3, sd = 0.4),   # Control 1
+    rnorm(N1, mean = 3.5, sd = 0.5), # Test 1
+    rnorm(N2, mean = 2.5, sd = 0.4), # Control 2
+    rnorm(N2, mean = 3.0, sd = 0.5)  # Test 2
+  ),
+  ID = c(1:N1, 1:N1, 1:N2, 1:N2)
+)
+```
+
+```{r, warning = FALSE}
+# Each pair is internally balanced even though the two pairs differ in size.
+multi_unequal_pairs.mean_diff <- load(df_unequal_pairs,
+  x = Group, y = Measurement,
+  idx = list(c("Control 1", "Test 1"), c("Control 2", "Test 2")),
+  paired = "sequential", id_col = ID
+) %>%
+  mean_diff()
+
+dabest_plot(multi_unequal_pairs.mean_diff,
+  raw_marker_size = 0.5, raw_marker_alpha = 0.3
+)
+```
+
+An unbalanced pair (different n *within* a pair) is still caught as an error:
+
+```{r, error = TRUE}
+# Remove one row from Test 1 so Control 1 (n=20) and Test 1 (n=19) are mismatched within the pair.
+df_bad_pair <- df_unequal_pairs %>%
+  dplyr::filter(!(Group == "Test 1" & ID == 1))
+
+load(df_bad_pair,
+  x = Group, y = Measurement,
+  idx = list(c("Control 1", "Test 1"), c("Control 2", "Test 2")),
+  paired = "sequential", id_col = ID
+)
+```