Rename RMSA to RMSR in model_performance() for psych FA objects (#884)

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: DominiqueMakowski <8875533+DominiqueMakowski@users.noreply.github.com>
Co-authored-by: Daniel <mail@danielluedecke.de>
Co-authored-by: strengejacke <26301769+strengejacke@users.noreply.github.com>

Author: 3 people

NEWS.md

@@ -2,6 +2,10 @@
 
 ## Changes
 
+* `model_performance()` for psych FA objects now correctly names the metric as 
+  `RMSR` (Root Mean Square Residual) instead of `RMSA`. The `RMSR_corrected` 
+  column (previously `RMSA_corrected`) is also renamed accordingly.
+
 * `check_model()` now limits the number of data points for models with many
   observations, to reduce the time for rendering the plot via the `maximum_dots`
   argument.

R/model_performance.psych.R

@@ -7,13 +7,17 @@
 #' `principal` (e.g., from `psych::principal()`), or from
 #' `parameters::factor_analysis()` or `item_omega()`.
 #' @param metrics Can be `"all"` or a character vector of metrics to be computed
-#' (some of `"Chi2"`, `"Chi2_df"`, `"df"`, `"p_Chi2"`, `"RMSA"`,
-#' `"RMSA_corrected"`, `"TLI"`, `"RMSEA"`, and `"BIC"`. For omega-models, can
+#' (some of `"Chi2"`, `"Chi2_df"`, `"df"`, `"p_Chi2"`, `"RMSR"`,
+#' `"RMSR_corrected"`, `"TLI"`, `"RMSEA"`, and `"BIC"`. For omega-models, can
 #' also include `"R2"` and `"Correlation"`.
 #' @param verbose Toggle off warnings.
 #' @param ... Arguments passed to or from other methods.
 #'
 #' @details
+#' The `RMSR_corrected` metric is the Root Mean Square Residual corrected for
+#' degrees of freedom (i.e., dividing by degrees of freedom rather than the
+#' number of observations).
+#'
 #' For omega-models, the columns `R2` and `Correlation` are measures of factor
 #' score adequacy. `R2` refers to the multiple R square of scores with factors,
 #' while `Correlation` indicates the correlation of scores with factors.
@@ -33,8 +37,8 @@ model_performance.fa <- function(model, metrics = "all", verbose = TRUE, ...) {
     Chi2 = ifelse(is.null(model$STATISTIC), NA_real_, model$STATISTIC),
     Chi2_df = ifelse(is.null(model$dof), NA_real_, model$dof),
     p_Chi2 = ifelse(is.null(model$PVAL), NA_real_, model$PVAL),
-    RMSA = ifelse(is.null(model$rms), NA_real_, model$rms),
-    RMSA_corrected = ifelse(is.null(model$crms), NA_real_, model$crms),
+    RMSR = ifelse(is.null(model$rms), NA_real_, model$rms),
+    RMSR_corrected = ifelse(is.null(model$crms), NA_real_, model$crms),
     TLI = ifelse(is.null(model$TLI), NA_real_, model$TLI),
     RMSEA = ifelse(is.null(model$RMSEA), NA_real_, model$RMSEA[1]),
     RMSEA_CI = ifelse(is.null(model$RMSEA), NA_real_, 0.9),
@@ -92,8 +96,8 @@ model_performance.omega <- function(model, metrics = "all", verbose = TRUE, ...)
         Chi2 = ifelse(is.null(stats$STATISTIC), NA_real_, stats$STATISTIC),
         df = ifelse(is.null(stats$dof), NA_real_, stats$dof),
         p_Chi2 = ifelse(is.null(stats$PVAL), NA_real_, stats$PVAL),
-        RMSA = ifelse(is.null(stats$rms), NA_real_, stats$rms),
-        RMSA_corrected = ifelse(is.null(stats$crms), NA_real_, stats$crms),
+        RMSR = ifelse(is.null(stats$rms), NA_real_, stats$rms),
+        RMSR_corrected = ifelse(is.null(stats$crms), NA_real_, stats$crms),
         TLI = ifelse(is.null(stats$TLI), NA_real_, stats$TLI),
         RMSEA = ifelse(is.null(stats$RMSEA), NA_real_, stats$RMSEA[1]),
         RMSEA_CI = ifelse(is.null(stats$RMSEA), NA_real_, 0.9),
@@ -148,7 +152,7 @@ print.performance_omega <- function(x, ...) {
   n <- attr(x, "n", exact = TRUE)
   insight::print_color(
     insight::format_message(sprintf(
-      "\nCompare the model fit of the %i-factor solution with the g-only model. If the g-model has smaller RMSA and RMSEA then your items are more likely to describe a single unidimensional construct. If the %i-factor model has smaller RMSA and RMSEA then your construct is more likely to be made up of %i sub-constructs.",
+      "\nCompare the model fit of the %i-factor solution with the g-only model. If the g-model has smaller RMSR and RMSEA then your items are more likely to describe a single unidimensional construct. If the %i-factor model has smaller RMSR and RMSEA then your construct is more likely to be made up of %i sub-constructs.",
       n,
       n,
       n

man/model_performance.fa.Rd

@@ -89,7 +89,7 @@
     Output
       # Indices of model performance
       
-      Chi2(167) | p (Chi2) |  RMSA | RMSA_corrected |   TLI | RMSEA |   RMSEA 90% CI |     BIC
+      Chi2(167) | p (Chi2) |  RMSR | RMSR_corrected |   TLI | RMSEA |   RMSEA 90% CI |     BIC
       ----------------------------------------------------------------------------------------
       267.210   |   < .001 | 0.013 |          0.015 | 0.991 | 0.015 | [0.012, 0.019] | -1044.1
 
@@ -100,7 +100,7 @@
     Output
       # Indices of model performance
       
-      Chi2(1) | p (Chi2) |  RMSA
+      Chi2(1) | p (Chi2) |  RMSR
       --------------------------
       4.119   |    0.042 | 0.035
 
@@ -111,15 +111,15 @@
     Output
       # Indices of model performance
       
-      Model             |    Chi2 | df | p (Chi2) |  RMSA | RMSA_corrected |   TLI | RMSEA |   RMSEA 90% CI |   BIC |    R2 | Correlation
+      Model             |    Chi2 | df | p (Chi2) |  RMSR | RMSR_corrected |   TLI | RMSEA |   RMSEA 90% CI |   BIC |    R2 | Correlation
       -----------------------------------------------------------------------------------------------------------------------------------
       3-factor solution |  31.796 | 25 |   0.164  | 0.015 |          0.023 |       | 0.087 | [0.000, 0.181] | -54.8 |       |            
       g-model           | 264.781 | 44 |   < .001 | 0.393 |          0.440 | 0.195 | 0.395 | [0.356, 0.450] | 112.3 | 0.761 |       0.873
       
       Compare the model fit of the 3-factor solution with the g-only model.
-        If the g-model has smaller RMSA and RMSEA then your items are more
+        If the g-model has smaller RMSR and RMSEA then your items are more
         likely to describe a single unidimensional construct. If the 3-factor
-        model has smaller RMSA and RMSEA then your construct is more likely to
+        model has smaller RMSR and RMSEA then your construct is more likely to
         be made up of 3 sub-constructs.
 
 ---
@@ -129,14 +129,14 @@
     Output
       # Indices of model performance
       
-      Model             |    Chi2 | df | p (Chi2) |  RMSA | RMSA_corrected |   TLI | RMSEA |   RMSEA 90% CI |   BIC |    R2 | Correlation
+      Model             |    Chi2 | df | p (Chi2) |  RMSR | RMSR_corrected |   TLI | RMSEA |   RMSEA 90% CI |   BIC |    R2 | Correlation
       -----------------------------------------------------------------------------------------------------------------------------------
       3-factor solution |  31.796 | 25 |   0.164  | 0.015 |          0.023 |       | 0.087 | [0.000, 0.181] | -54.8 |       |            
       g-model           | 264.781 | 44 |   < .001 | 0.393 |          0.440 | 0.195 | 0.395 | [0.356, 0.450] | 112.3 | 0.761 |       0.873
       
       Compare the model fit of the 3-factor solution with the g-only model.
-        If the g-model has smaller RMSA and RMSEA then your items are more
+        If the g-model has smaller RMSR and RMSEA then your items are more
         likely to describe a single unidimensional construct. If the 3-factor
-        model has smaller RMSA and RMSEA then your construct is more likely to
+        model has smaller RMSR and RMSEA then your construct is more likely to
         be made up of 3 sub-constructs.