Merge pull request #622 from easystats/iss620

#620

DESCRIPTION

@@ -71,7 +71,7 @@ Depends:
 Imports:
     bayestestR (>= 0.11.0),
     datawizard (>= 0.2.1),
-    insight (>= 0.14.4),
+    insight (>= 0.14.5),
     graphics,
     methods,
     stats,
@@ -101,7 +101,7 @@ Suggests:
     dplyr,
     drc,
     DRR,
-    effectsize (>= 0.5),
+    effectsize (>= 0.5.0.1),
     EGAnet (>= 0.7),
     emmeans (>= 1.7.0),
     factoextra,
@@ -174,6 +174,8 @@ Suggests:
     truncreg,
     VGAM,
     WRS2
+Remotes:
+    easystats/effectsize
 VignetteBuilder: 
     knitr
 Encoding: UTF-8

NEWS.md

@@ -6,12 +6,17 @@
 
 ## Changes to functions
 
-* `model_parameters()` for `merMod` models now also computes CIs for the
-  random SD parameters when `ci_method="boot"` (previously, this was only
-  possible when `ci_method` was `"profile"`).
+* `model_parameters()` for `merMod` models now also computes CIs for the random
+  SD parameters when `ci_method="boot"` (previously, this was only possible when
+  `ci_method` was `"profile"`).
 
-* `model_parameters()` for `glmmTMB` models now computes CIs for the
-  random SD parameters. Note that these are based on a Wald-z-distribution.
+* `model_parameters()` for `glmmTMB` models now computes CIs for the random SD
+  parameters. Note that these are based on a Wald-z-distribution.
+
+* Similar to `model_parameters.htest()`, the `model_parameters.BFBayesFactor()`
+  method gains `cohens_d` and `cramers_v` arguments to control if you need to
+  add frequentist effect size estimates to the returned summary dataframe.
+  Previously, this was done by default.
 
 ## Bug fixes 
 
@@ -56,13 +61,14 @@ necessary changes before this breaking change is implemented.
 
 ## General
 
-* The handling to approximate the degrees of freedom in `model_parameters()`, 
+* The handling to approximate the degrees of freedom in `model_parameters()`,
   `ci()` and `p_value()` was revised and should now be more consistent. Some
   bugs related to the previous computation of confidence intervals and p-values
-  have been fixed. Now it is possible to change the method to approximate 
+  have been fixed. Now it is possible to change the method to approximate
   degrees of freedom for CIs and p-values using the `ci_method`, resp. `method`
-  argument. This change has been documented in detail in `?model_parameters`, 
-  and online here: https://easystats.github.io/parameters/reference/model_parameters.html
+  argument. This change has been documented in detail in `?model_parameters`,
+  and online here:
+  https://easystats.github.io/parameters/reference/model_parameters.html
 
 * Minor changes to `print()` for *glmmTMB* with dispersion parameter.
 
@@ -82,8 +88,8 @@ necessary changes before this breaking change is implemented.
   option for the `ci_method` (resp. `method`) argument, to explicitly calculate
   confidence intervals based on the residual degrees of freedom, when present.
 
-* `model_parameters()` supports following new objects:
-  `trimcibt`, `wmcpAKP`, `dep.effect` (in *WRS2* package), `systemfit`
+* `model_parameters()` supports following new objects: `trimcibt`, `wmcpAKP`,
+  `dep.effect` (in *WRS2* package), `systemfit`
 
 * `model_parameters()` gains a new argument `table_wide` for ANOVA tables. This
   can be helpful for users who may wish to report ANOVA table in wide format
@@ -99,19 +105,19 @@ necessary changes before this breaking change is implemented.
   not the default value, the printed output includes a message indicating which
   approximation-method for degrees of freedom was used.
 
-* `model_parameters()` for mixed models with `ci_method = "profile` computes 
-  (profiled) confidence intervals for both fixed and random effects. Thus, 
-  `ci_method = "profile` allows to add confidence intervals to the random
-  effect variances.
+* `model_parameters()` for mixed models with `ci_method = "profile` computes
+  (profiled) confidence intervals for both fixed and random effects. Thus,
+  `ci_method = "profile` allows to add confidence intervals to the random effect
+  variances.
 
 * `model_parameters()` should longer fail for supported model classes when
   robust standard errors are not available.
 
 ### Other functions
 
-* `n_factors()` the methods based on fit indices have been fixed and can be 
-  included separately (`package = "fit"`). Also added a `n_max` argument to 
-  crop the output.  
+* `n_factors()` the methods based on fit indices have been fixed and can be
+  included separately (`package = "fit"`). Also added a `n_max` argument to crop
+  the output.
 
 * `compare_parameters()` now also accepts a list of model objects.
 
@@ -121,9 +127,9 @@ necessary changes before this breaking change is implemented.
 * `format_parameters()` removes dots and underscores from parameter names, to
   make these more "human readable".
 
-* The experimental calculation of p-values in `equivalence_test()` was 
-  replaced by a proper calculation p-values. The argument `p_value` was removed
-  and p-values are now always included.
+* The experimental calculation of p-values in `equivalence_test()` was replaced
+  by a proper calculation p-values. The argument `p_value` was removed and
+  p-values are now always included.
 
 * Minor improvements to `print()`, `print_html()` and `print_md()`.
 
@@ -139,7 +145,7 @@ necessary changes before this breaking change is implemented.
   returned residual degrees of freedom in the statistic column, but confidence
   intervals were based on `Inf` degrees of freedom instead.
 
-* Fixed issue in `ci_satterthwaite()`, which used `Inf` degrees of freedom 
+* Fixed issue in `ci_satterthwaite()`, which used `Inf` degrees of freedom
   instead of the Satterthwaite approximation.
 
 * Fixed issue in `model_parameters.mlm()` when model contained interaction
@@ -151,8 +157,8 @@ necessary changes before this breaking change is implemented.
 * Fixed sign error for `model_parameters.htest()` for objects created with
   `t.test.formula()` (issue #552)
 
-* Fixed issue when computing random effect variances in `model_parameters()` 
-  for mixed models with categorical random slopes.
+* Fixed issue when computing random effect variances in `model_parameters()` for
+  mixed models with categorical random slopes.
 
 # parameters 0.14.0
 

R/1_model_parameters.R

@@ -395,7 +395,6 @@ parameters <- model_parameters
 #' # different p-value style in output
 #' model_parameters(model, p_digits = 5)
 #' model_parameters(model, digits = 3, ci_digits = 4, p_digits = "scientific")
-#'
 #' \donttest{
 #' # logistic regression model
 #' model <- glm(vs ~ wt + cyl, data = mtcars, family = "binomial")

R/2_ci.R

@@ -68,7 +68,6 @@ ci.glm <- function(x,
                    method = "profile",
                    robust = FALSE,
                    ...) {
-
   method <- match.arg(method, choices = c("profile", "wald", "normal", "residual"))
   if (method == "profile") {
     out <- lapply(ci, function(i) .ci_profiled(model = x, ci = i))

R/check_heterogeneity.R

@@ -21,7 +21,6 @@
 #'
 #' @export
 check_heterogeneity <- function(x, select = NULL, group = NULL) {
-
   .Deprecated("performance::check_heterogeneity_bias()")
 
   if (insight::is_model(x)) {

R/ci_generic.R

@@ -12,9 +12,11 @@
     method <- "wald"
   }
   method <- tolower(method)
-  method <- match.arg(method, choices = c("wald", "ml1", "betwithin", "kr",
-                                          "satterthwaite", "kenward", "boot",
-                                          "profile", "residual", "normal"))
+  method <- match.arg(method, choices = c(
+    "wald", "ml1", "betwithin", "kr",
+    "satterthwaite", "kenward", "boot",
+    "profile", "residual", "normal"
+  ))
 
   effects <- match.arg(effects)
   component <- match.arg(component)

R/cluster_analysis.R

@@ -232,4 +232,3 @@ print.cluster_analysis <- function(x, digits = 2, ...) {
 
   list(model = model, clusters = model$cluster)
 }
-

R/dof.R

@@ -70,7 +70,6 @@ degrees_of_freedom <- function(model, ...) {
 #' @rdname degrees_of_freedom
 #' @export
 degrees_of_freedom.default <- function(model, method = "analytical", ...) {
-
   if (is.null(method)) {
     method <- "wald"
   }

R/format.R

@@ -656,22 +656,24 @@ format.parameters_distribution <- function(x, digits = 2, format = NULL, ci_widt
     }
 
     string_tailed <- switch(toupper(ci_method),
-                            "HDI" = "highest-density",
-                            "UNIROOT" = ,
-                            "PROFILE" = "profile-likelihood",
-                            "equal-tailed")
+      "HDI" = "highest-density",
+      "UNIROOT" = ,
+      "PROFILE" = "profile-likelihood",
+      "equal-tailed"
+    )
 
     string_method <- switch(toupper(ci_method),
-                            "BCI" = ,
-                            "BCAI" = "bias-corrected accelerated bootstrap",
-                            "SI" = ,
-                            "CI" = ,
-                            "QUANTILE" = ,
-                            "ETI" = ,
-                            "HDI" = ifelse(isTRUE(bootstrap), "na\u0131ve bootstrap", "MCMC"),
-                            "NORMAL" = "Wald normal",
-                            "BOOT" = "parametric bootstrap",
-                            "Wald")
+      "BCI" = ,
+      "BCAI" = "bias-corrected accelerated bootstrap",
+      "SI" = ,
+      "CI" = ,
+      "QUANTILE" = ,
+      "ETI" = ,
+      "HDI" = ifelse(isTRUE(bootstrap), "na\u0131ve bootstrap", "MCMC"),
+      "NORMAL" = "Wald normal",
+      "BOOT" = "parametric bootstrap",
+      "Wald"
+    )
 
     if (toupper(ci_method) %in% c("KENWARD", "KR", "KENWARD-ROGER", "KENWARD-ROGERS", "SATTERTHWAITE")) {
       string_approx <- paste0("with ", format_df_adjust(ci_method, approx_string = "", dof_string = ""), " ")
@@ -681,10 +683,11 @@ format.parameters_distribution <- function(x, digits = 2, format = NULL, ci_widt
 
     if (!is.null(test_statistic) && !ci_method %in% c("normal") && !isTRUE(bootstrap)) {
       string_statistic <- switch(tolower(test_statistic),
-                                 "t-statistic" = "t",
-                                 "chi-squared statistic" = ,
-                                 "z-statistic" = "z",
-                                 "")
+        "t-statistic" = "t",
+        "chi-squared statistic" = ,
+        "z-statistic" = "z",
+        ""
+      )
       string_method <- paste0(string_method, " ", string_statistic, "-")
     } else {
       string_method <- paste0(string_method, " ")

R/methods_BayesFactor.R

@@ -3,11 +3,15 @@
 
 #' Parameters from BayesFactor objects
 #'
-#' Parameters from BayesFactor objects.
+#' Parameters from `BFBayesFactor` objects from `{BayesFactor}` package.
 #'
 #' @param model Object of class `BFBayesFactor`.
+#' @param cohens_d If `TRUE`, compute Cohens' *d* as index of effect size. Only
+#'   applies to objects from `ttestBF()`. See `effectsize::cohens_d()` for
+#'   details.
 #' @inheritParams bayestestR::describe_posterior
 #' @inheritParams p_value
+#' @inheritParams model_parameters.htest
 #'
 #' @details
 #' The meaning of the extracted parameters:
@@ -30,6 +34,7 @@
 #' if (require("BayesFactor")) {
 #'   model <- ttestBF(x = rnorm(100, 1, 1))
 #'   model_parameters(model)
+#'   model_parameters(model, cohens_d = TRUE, ci = .9)
 #' }
 #' }
 #' @return A data frame of indices related to the model's parameters.
@@ -43,6 +48,8 @@ model_parameters.BFBayesFactor <- function(model,
                                            rope_range = "default",
                                            rope_ci = 0.95,
                                            priors = TRUE,
+                                           cohens_d = NULL,
+                                           cramers_v = NULL,
                                            verbose = TRUE,
                                            ...) {
   if (any(grepl("^Null", names(model@numerator)))) {
@@ -103,7 +110,11 @@ model_parameters.BFBayesFactor <- function(model,
   )
 
   # Effect size?
-  if (requireNamespace("effectsize", quietly = TRUE) && bf_type %in% c("ttest1", "ttest2", "xtable")) {
+  if (bf_type %in% c("ttest1", "ttest2") && !is.null(cohens_d) ||
+    bf_type == "xtable" && !is.null(cramers_v)) {
+    # needs {effectsize} to be installed
+    insight::check_if_installed("effectsize")
+
     tryCatch(
       {
         effsize <- effectsize::effectsize(model,

R/methods_betareg.R

@@ -44,7 +44,6 @@ ci.betareg <- function(x,
                        ci = .95,
                        component = "all",
                        ...) {
-
   component <- match.arg(component, choices = c("all", "conditional", "precision"))
   .ci_generic(model = x, ci = ci, dof = Inf, component = component)
 }

R/methods_dbscan.R

@@ -2,7 +2,6 @@
 #' @inheritParams cluster_centers
 #'
 #' @examples
-#'
 #' \donttest{
 #' # DBSCAN ---------------------------
 #' if (require("dbscan", quietly = TRUE)) {

R/methods_glmmTMB.R

@@ -190,7 +190,6 @@ ci.glmmTMB <- function(x,
                        component = "all",
                        verbose = TRUE,
                        ...) {
-
   method <- tolower(method)
   method <- match.arg(method, choices = c("wald", "normal", "ml1", "betwithin", "profile", "uniroot"))
   component <- match.arg(component, choices = c("all", "conditional", "zi", "zero_inflated", "dispersion"))

R/methods_htest.R

@@ -239,9 +239,22 @@ model_parameters.pairwise.htest <- function(model, verbose = TRUE, ...) {
     # chi2- and mcnemar-test -----------
     out <- .extract_htest_chi2(model)
     if (grepl("^McNemar", model$method)) {
-      out <- .add_effectsize_mcnemar(model, out, cohens_g = cohens_g, ci = ci, verbose = verbose)
+      out <- .add_effectsize_mcnemar(model,
+        out,
+        cohens_g = cohens_g,
+        ci = ci,
+        verbose = verbose
+      )
     } else {
-      out <- .add_effectsize_chi2(model, out, cramers_v = cramers_v, phi = phi, ci = ci, alternative = alternative, verbose = verbose)
+      out <- .add_effectsize_chi2(
+        model,
+        out,
+        cramers_v = cramers_v,
+        phi = phi,
+        ci = ci,
+        alternative = alternative,
+        verbose = verbose
+      )
     }
   } else if (m_info$is_proptest) {
 

R/methods_lme4.R

@@ -90,8 +90,8 @@ model_parameters.merMod <- function(model,
       ci_method <- "quantile"
     } else {
       ci_method <- switch(insight::find_statistic(model),
-                          "t-statistic" = "residual",
-                          "wald"
+        "t-statistic" = "residual",
+        "wald"
       )
     }
   }
@@ -236,11 +236,12 @@ ci.merMod <- function(x,
                       robust = FALSE,
                       iterations = 500,
                       ...) {
-
   method <- tolower(method)
-  method <- match.arg(method, choices = c("wald", "ml1", "betwithin", "kr",
-                                          "satterthwaite", "kenward", "boot",
-                                          "profile", "residual", "normal"))
+  method <- match.arg(method, choices = c(
+    "wald", "ml1", "betwithin", "kr",
+    "satterthwaite", "kenward", "boot",
+    "profile", "residual", "normal"
+  ))
 
   # bootstrapping
   if (method == "boot") {

R/methods_survival.R

@@ -38,7 +38,7 @@ p_value.coxph <- function(model, ...) {
 
   # convert in case of z
   if (identical(statistic, "z-statistic")) {
-    params$Statistic <- params$Statistic ^ 2
+    params$Statistic <- params$Statistic^2
   }
 
   .data_frame(