construct_model() updates

NEWS.md

@@ -24,29 +24,30 @@ ard_moodtest()          -> ard_stats_mood_test()
   - `ard_stats_mcnemar_test_long()` for McNemar's test from long data using `stats::mcnemar.test()`. 
   - `ard_aod_wald_test()` for calculating Wald Tests for regression models using `aod::wald.test()`. (#84)
   - `ard_car_anova()` for calculating ANOVA results using `car::Anova()`. (#3)
+  - `ard_car_vif()` for calculating the variance inflation factor using `car::vif()`. (#10)
   - `ard_stats_oneway_test()` for calculating ANOVA results using `stats::oneway.test()`. (#3)
   - `ard_effectsize_cohens_d()`, `ard_effectsize_paired_cohens_d()`, `ard_effectsize_hedges_g()`, and `ard_effectsize_paired_hedges_g()` for standardized differences using `effectsize::cohens_d()` and `effectsize::hedges_g()`. (#50)
-  - `ard_stats_prop_test()` for tests of proportions using `stats::prop.test()`. (#64)
-  - `ard_regression_basic()` for basic regression models. The function focuses on matching terms to underlying variables names. (#46)
+  - `ard_emmeans_mean_difference()` for calculating the least-squares mean differences using the {emmeans} package. (#34)
   - `ard_smd_smd()` for calculating standardized mean differences using `smd::smd()`. (#4)
   - `ard_survival_survfit()` for survival analyses using `survival::survfit()`. (#43)
   - `ard_continuous.survey.design()` for calculating univariate summary statistics from weighted/survey data using many functions from the {survey} package. (#68)
   - `ard_attributes.survey.design()` for summarizing labels and attributes from weighted/survey data using many functions from the {survey} package.
   - `ard_survey_svychisq()` for weighted/survey chi-squared test using `survey::svychisq()`. (#72)
   - `ard_survey_svyttest()` for weighted/survey t-tests using `survey::svyttest()`. (#70)
   - `ard_survey_svyranktest()` for weighted/survey rank tests using `survey::svyranktest()`. (#71)
-  - `ard_car_vif()` for calculating the variance inflation factor using `car::vif()`. (#10)
-  - `ard_emmeans_mean_difference()` for calculating the least-squares mean differences using the {emmeans} package. (#34)
-  - `ard_stats_wilcox_test_onesample()` for calculating one-sample results.
+  - `ard_survival_survdiff()` for creating results from `survival::survdiff()`. (#113)
+  - `ard_stats_prop_test()` for tests of proportions using `stats::prop.test()`. (#64)
   - `ard_stats_t_test_onesample()` for calculating one-sample results.
+  - `ard_stats_wilcox_test_onesample()` for calculating one-sample results.
+  - `ard_regression_basic()` for basic regression models. The function focuses on matching terms to underlying variables names. (#46)
 
 * Updated functions `ard_stats_t_test()`, `ard_stats_paired_t_test()`, `ard_stats_wilcox_test()`, `ard_stats_paired_wilcox_test()`, `ard_stats_chisq_test()`, `ard_stats_fisher_test()`, `ard_stats_kruskal_test()`, `ard_stats_mcnemar_test()`, and `ard_stats_mood_test()` to accept multiple variables at once. Independent tests are calculated for each variable. The `variable` argument is renamed to `variables`. (#77)
 
 * Updated `ard_stats_t_test()` and `ard_stats_wilcox_test()` to no longer require the `by` argument, which yields central estimates with their confidence intervals. (#82)
 
 * Imported cli call environment functions from `https://github.com/ddsjoberg/standalone/blob/main/R/standalone-cli_call_env.R` and implemented `set_cli_abort_call` in user-facing functions. (#111)
 
-* Added `ard_survival_survdiff()` for creating results from `survival::survdiff()`. (#113)
+* Added model construction helpers, `construct_model()`, `reformulate2()`, `bt()`, and `bt_strip()`.
 
 # cardx 0.1.0
 

R/ard_emmeans_mean_difference.R

@@ -70,7 +70,7 @@ ard_emmeans_mean_difference <- function(data, formula, method,
   # construct primary model ----------------------------------------------------
   mod <-
     construct_model(
-      x = data, formula = formula, method = method,
+      data = data, formula = formula, method = method,
       method.args = {{ method.args }},
       package = package, env = caller_env()
     )

R/ard_stats_anova.R

@@ -122,7 +122,7 @@ ard_stats_anova.data.frame <- function(x,
         lapply(
           formulas,
           function(formula) {
-            construct_model(x = x, formula = formula, method = method, method.args = {{ method.args }}, package = package)
+            construct_model(data = x, formula = formula, method = method, method.args = {{ method.args }}, package = package)
           }
         )
 

R/construction_helpers.R

@@ -13,14 +13,15 @@
 #'
 #' - `bt_strip()`: Removes backticks from a string if it begins and ends with a backtick.
 #'
-#' @param x
+#' @param data
 #'  - `construct_model.data.frame()` (`data.frame`) a data frame
 #'  - `construct_model.survey.design()` (`survey.design`) a survey design object
-#'  - `bt()`/`bt_strip()` (`character`) character vector, typically of variable names
+#' @param x (`character`)\cr
+#'  character vector, typically of variable names
 #' @param formula (`formula`)\cr
 #'   a formula
 #' @param method (`string`)\cr
-#'   string naming the function to be called, e.g. `"glm"`.
+#'   string of function naming the function to be called, e.g. `"glm"`.
 #'   If function belongs to a library that is not attached, the package name
 #'   must be specified in the `package` argument.
 #' @param method.args (named `list`)\cr
@@ -47,7 +48,7 @@
 #'
 #' @examplesIf do.call(asNamespace("cardx")$is_pkg_installed, list(pkg = c("withr", "lme4", "broom.helpers", "broom.mixed"), reference_pkg = "cardx"))
 #' construct_model(
-#'   x = mtcars,
+#'   data = mtcars,
 #'   formula = am ~ mpg + (1 | vs),
 #'   method = "glmer",
 #'   method.args = list(family = binomial),
@@ -56,7 +57,7 @@
 #'   broom.mixed::tidy()
 #'
 #' construct_model(
-#'   x = mtcars |> dplyr::rename(`M P G` = mpg),
+#'   data = mtcars |> dplyr::rename(`M P G` = mpg),
 #'   formula = reformulate2(c("M P G", "cyl"), response = "hp"),
 #'   method = "lm"
 #' ) |>
@@ -66,13 +67,13 @@ NULL
 
 #' @rdname construction_helpers
 #' @export
-construct_model <- function(x, ...) {
+construct_model <- function(data, ...) {
   UseMethod("construct_model")
 }
 
 #' @rdname construction_helpers
 #' @export
-construct_model.data.frame <- function(x, formula, method, method.args = list(), package = "base", env = caller_env(), ...) {
+construct_model.data.frame <- function(data, formula, method, method.args = list(), package = "base", env = caller_env(), ...) {
   set_cli_abort_call()
   # check pkg installations ----------------------------------------------------
   check_dots_empty()
@@ -82,23 +83,23 @@ construct_model.data.frame <- function(x, formula, method, method.args = list(),
   check_class(formula, cls = "formula")
 
   check_not_missing(method)
-  check_string(method)
-  check_not_namespaced(method)
+  check_string_or_function(method)
+  if (is_string(method)) check_not_namespaced(method)
 
   # convert method.args to list of expressions (to account for NSE inputs) -----
   method.args <- .as_list_of_exprs({{ method.args }})
 
   # build model ----------------------------------------------------------------
   withr::with_namespace(
     package = package,
-    call2(.fn = method, formula = formula, data = x, !!!method.args) |>
+    call2(.fn = method, formula = formula, data = data, !!!method.args) |>
       eval_tidy(env = env)
   )
 }
 
 #' @rdname construction_helpers
 #' @export
-construct_model.survey.design <- function(x, formula, method, method.args = list(), package = "survey", env = caller_env(), ...) {
+construct_model.survey.design <- function(data, formula, method, method.args = list(), package = "survey", env = caller_env(), ...) {
   set_cli_abort_call()
   # check pkg installations ----------------------------------------------------
   check_dots_empty()
@@ -108,7 +109,7 @@ construct_model.survey.design <- function(x, formula, method, method.args = list
   check_class(formula, cls = "formula")
 
   check_not_missing(method)
-  check_string(method)
+  check_string_or_function(method)
   check_not_namespaced(method)
 
   # convert method.args to list of expressions (to account for NSE inputs) -----
@@ -117,7 +118,7 @@ construct_model.survey.design <- function(x, formula, method, method.args = list
   # build model ----------------------------------------------------------------
   withr::with_namespace(
     package = package,
-    call2(.fn = method, formula = formula, design = x, !!!method.args) |>
+    call2(.fn = method, formula = formula, design = data, !!!method.args) |>
       eval_tidy(env = env)
   )
 }
@@ -176,10 +177,27 @@ check_not_namespaced <- function(x,
   check_string(x, arg_name = arg_name, call = call, class = "check_not_namespaced")
 
   if (str_detect(x, "::")) {
-    c("Argument {.arg {arg_name}} cannot be namespaced.",
-      i = "Put the package name in the {.arg package} argument."
-    ) |>
-      cli::cli_abort(call = call, class = class)
+    cli::cli_abort(
+      "Argument {.arg {arg_name}} cannot be namespaced when passed as a {.cls string}.",
+      call = call,
+      class = class
+    )
+  }
+
+  invisible(x)
+}
+
+
+check_string_or_function <- function(x,
+                                     arg_name = rlang::caller_arg(x),
+                                     class = "check_string_or_function",
+                                     call = get_cli_abort_call()) {
+  if (!is.function(x) && !is_string(x)) {
+    cli::cli_abort(
+      c("Argument {.arg {arg_name}} must be a {.cls string} or {.cls function}."),
+      call = call,
+      class = class
+    )
   }
 
   invisible(x)

README.Rmd

@@ -74,7 +74,7 @@ To accomplish this we include model construction helpers.
 
 ```{r}
 construct_model(
-  x = cards::ADSL,
+  data = cards::ADSL,
   formula = reformulate2("ARM", response = "AGE"),
   method = "lm"
 ) |>

README.md

@@ -95,7 +95,7 @@ this we include model construction helpers.
 
 ``` r
 construct_model(
-  x = cards::ADSL,
+  data = cards::ADSL,
   formula = reformulate2("ARM", response = "AGE"),
   method = "lm"
 ) |>

tests/testthat/_snaps/construction_helpers.md

@@ -1,7 +1,7 @@
 # construct_model() works
 
     Code
-      dplyr::filter(as.data.frame(ard_regression(construct_model(x = dplyr::rename(
+      dplyr::filter(as.data.frame(ard_regression(construct_model(data = dplyr::rename(
         mtcars, `M P G` = mpg), formula = reformulate2(c("M P G", "cyl"), response = "hp"),
       method = "lm"))), stat_name %in% c("term", "estimate", "p.value"))
     Output
@@ -13,3 +13,20 @@
       5      cyl regression  estimate Coefficient    23.97863      1
       6      cyl regression   p.value     p-value 0.002814958      1
 
+# construct_model() messaging
+
+    Code
+      construct_model(data = mtcars, method = "survival::coxph", formula = survival::Surv(
+        mpg, am) ~ cyl)
+    Condition
+      Error in `construct_model()`:
+      ! Argument `method` cannot be namespaced when passed as a <string>.
+
+---
+
+    Code
+      construct_model(data = mtcars, method = letters, formula = am ~ cyl)
+    Condition
+      Error in `construct_model()`:
+      ! Argument `method` must be a <string> or <function>.
+