Use internal copy of to_numeric

Since it will be removed from datawizard: easystats/datawizard#197 (comment)

modelbased needs to be updated on CRAN before datawizard.

R/estimate_grouplevel.R

@@ -83,7 +83,7 @@ estimate_grouplevel <- function(model, type = "random", ...) {
   random <- random[c("Group", "Level", names(random)[!names(random) %in% c("Group", "Level")])]
 
   # Sort
-  random <- random[order(random$Group, datawizard::to_numeric(random$Level), random$Parameter), ]
+  random <- random[order(random$Group, .to_numeric(random$Level), random$Parameter), ]
 
   # Clean
   row.names(random) <- NULL

R/utils.R

@@ -0,0 +1,8 @@
+#' @keywords internal
+#' @noRd
+.to_numeric <- function(x) {
+  tryCatch(as.numeric(as.character(x)),
+    error = function(e) x,
+    warning = function(w) x
+  )
+}

R/visualisation_recipe.estimate_grouplevel.R

@@ -38,7 +38,7 @@ visualisation_recipe.estimate_grouplevel <- function(x,
                                                      ...) {
   data <- as.data.frame(x)
   # Fix order so that it's plotted with sorted levels
-  data$Level <- factor(data$Level, levels = sort(datawizard::to_numeric(unique(data$Level))))
+  data$Level <- factor(data$Level, levels = sort(.to_numeric(unique(data$Level))))
 
   layers <- list()