add initial tests

R/get_marginal_effect.R

@@ -60,6 +60,10 @@
 #'  marginal_se                \tab Standard error estimate of the marginal treatment effect estimate. \cr
 #'  marginal_results           \tab Analysis results data (ARD) containing a summary of the analysis for subsequent reporting. \cr
 #' }
+#' Note, the returned object is also appended with a `beeca` object class. This enables additional
+#' functions to support extracting the marginal effect estimates from the object without
+#' interacting with generic functions associated with the glm object.
+#'
 #' @importFrom utils packageVersion
 #' @export
 #' @examples

R/tidy_beeca.R

@@ -1,8 +1,8 @@
-#' Tidy a beeca object
+#' Tidy a beeca/glm object
 #'
 #' The broom package exports a tidier for `"glm"` objects.
 #' This function adds on top of that and returns more information
-#' that is stored from gcomputation.
+#' that is stored from g-computation.
 #'
 #' The function also utilizes additional information stored when the
 #' glm object is created with `get_marginal_effect()`.

tests/testthat/test-tidy_beeca.R

@@ -1,3 +1,49 @@
-test_that("multiplication works", {
-  expect_equal(2 * 2, 4)
+# set up example ----------------------------------------------------------
+data <- trial01
+data$trtp <- factor(data$trtp)
+data_complete <- na.omit(data)
+
+## Fit a logistic regression working model and pass it to beeca
+fit <- glm(aval ~ trtp + bl_cov, family = binomial(link = "logit"), data = data_complete) |>
+  get_marginal_effect(trt="trtp", method="Ye", contrast="diff", reference = "0")
+
+
+test_that("tidy_beeca stops if input is not a 'beeca' object", {
+  non_beeca_object <- lm(mpg ~ wt, data = mtcars)
+  expect_error(tidy_beeca(non_beeca_object), "x must be a 'beeca' object")
 })
+
+
+test_that("tidy_beeca returns a tibble", {
+  result <- tidy_beeca(fit)
+  expect_true(tibble::is_tibble(result))
+})
+
+test_that("tidy_beeca returns correct columns without confidence intervals", {
+  result <- tidy_beeca(fit, conf.int = FALSE)
+  expect_named(result, c("term", "contrast", "estimate", "std.error", "statistic", "p.value"))
+})
+
+test_that("tidy_beeca returns correct columns with confidence intervals", {
+  result <- tidy_beeca(fit, conf.int = TRUE)
+  expect_named(result, c("term", "contrast", "estimate", "std.error", "statistic", "p.value", "conf.low", "conf.high"))
+})
+
+test_that("tidy_beeca computes confidence intervals correctly", {
+  result <- tidy_beeca(fit, conf.int = TRUE, conf.level = 0.95)
+  conf.low <- result$estimate - qnorm(1 - 0.95 / 2) * result$std.error
+  conf.high <- result$estimate + qnorm(1 - 0.95 / 2) * result$std.error
+  expect_equal(result$conf.low, conf.low, tolerance = 1e-8)
+  expect_equal(result$conf.high, conf.high, tolerance = 1e-8)
+})
+
+
+test_that("tidy_beeca returns correct values", {
+  result <- tidy_beeca(fit, conf.int = FALSE)
+  expect_equal(result$term, "trtp")
+  expect_equal(result$estimate, fit$marginal_results$STATVAL[fit$marginal_results$STAT == "diff"])
+  expect_equal(result$std.error, fit$marginal_results$STATVAL[fit$marginal_results$STAT == "diff_se"])
+  expect_equal(result$statistic, NA)
+  expect_equal(result$p.value, NA)
+})
+