Preserve order of categories in the plot by elinw · Pull Request #448 · easystats/see

elinw · 2026-07-06T20:41:33Z

I noticed that when the category names are not in alphabetical order, ggplot2 alphabetizes when defining the x axis. This matters because sometimes order is important (in the case I noticed they defined a Likert scale). From what I learned doing some reading the way to address this is to turn the Category column into a factor. THat's what this PR does. That may have other consequences, I'm not sure yet.

Remove unneeded addition

gemini-code-assist

Code Review

This pull request introduces new S3 plot methods for grouped means (plot.dw_groupmeans and plot.dw_groupmeans_list) and crosstabs (plot.datawizard_crosstab), along with corresponding documentation and test suites. The code review highlights several critical and high-severity issues, including an undefined variable t_base in the crosstab plot function, a syntax error in the documentation examples, and compatibility issues with the native pipe operator |> on older R versions. Additionally, the feedback suggests improvements to confidence interval limit calculations (handling NA and negative values), ensuring category ordering is preserved in list plots, avoiding explicit print() calls in S3 plot methods, and correcting documentation tags.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

gemini-code-assist · 2026-07-06T20:44:32Z

+    x_long <- datawizard::data_to_long(
+      x,
+      rows_to = "row_var",
+      select = names(t_base)[-1]


The variable t_base is not defined anywhere in this function or file. This will cause an object 't_base' not found error at runtime. It should be x instead.

select = names(x)[-1]

This all was included by accident see new pR

gemini-code-assist · 2026-07-06T20:44:33Z

+#' group_means_object <- datawizard::means_by_group(
+#'   iris$Sepal.Width, iris$Species)
+#' )


There is a syntax error in the examples section due to an extra closing parenthesis on line 31. This will cause an error when running the examples or during package checks.

#' group_means_object <- datawizard::means_by_group(\n#' iris$Sepal.Width, iris$Species\n#' )

gemini-code-assist · 2026-07-06T20:44:33Z

+  if ("CI_low" %in% names(trimmed)) {
+    lower_lim <- .9 * min(trimmed$CI_low)
+    upper_lim <- 1.1 * max(trimmed$CI_high)
+  }


The current calculation of lower_lim and upper_lim has several issues:\n1. It does not handle NA values, which will cause min() and max() to return NA unless na.rm = TRUE is specified.\n2. It breaks for negative values (e.g., multiplying a negative minimum by 0.9 makes it larger/closer to zero, which truncates the lower part of the confidence interval).\n\nA more robust approach is to calculate the range of the confidence intervals and expand the limits relative to that range.

if ("CI_low" %in% names(trimmed)) {\n ci_range <- max(trimmed$CI_high, na.rm = TRUE) - min(trimmed$CI_low, na.rm = TRUE)\n lower_lim <- min(trimmed$CI_low, na.rm = TRUE) - 0.1 * ci_range\n upper_lim <- max(trimmed$CI_high, na.rm = TRUE) + 0.1 * ci_range\n }

This is wrong since these are never NA.

gemini-code-assist · 2026-07-06T20:44:33Z

+
+  x_long <- do.call(rbind, x)
+
+  trimmed <- datawizard::data_filter(x_long, Category != "Total")


In plot.dw_groupmeans_list, trimmed$Category is not converted to a factor. This means the main objective of this PR (preserving the order of categories in the plot) will not be applied when plotting a list of group means. We should convert trimmed$Category to a factor here as well, using unique(trimmed$Category) to handle the duplicate categories across different variables.

trimmed <- datawizard::data_filter(x_long, Category != "Total")\n trimmed$Category <- factor(trimmed$Category, levels = unique(trimmed$Category))

gemini-code-assist · 2026-07-06T20:44:33Z

+  if ("CI_low" %in% names(trimmed)) {
+    lower_lim <- .9 * min(trimmed$CI_low)
+    upper_lim <- 1.1 * max(trimmed$CI_high)
+  }


Just like in plot.dw_groupmeans, the calculation of lower_lim and upper_lim here does not handle NA values and breaks for negative values. Let's use the same robust range-based expansion.

if ("CI_low" %in% names(trimmed)) {\n ci_range <- max(trimmed$CI_high, na.rm = TRUE) - min(trimmed$CI_low, na.rm = TRUE)\n lower_lim <- min(trimmed$CI_low, na.rm = TRUE) - 0.1 * ci_range\n upper_lim <- max(trimmed$CI_high, na.rm = TRUE) + 0.1 * ci_range\n }

This is wrong.

gemini-code-assist · 2026-07-06T20:44:33Z

+#' (if present in the object should be graphed), and `caption` for if
+#' a caption summarizing ANOVA results should be included.
+#'
+#' @value


In roxygen2, the standard tag to document the return value of a function is @return, not @value. Using @value may prevent the return value from being correctly parsed and documented.

#' @return

gemini-code-assist · 2026-07-06T20:44:33Z

+    caption_text = ""
+  }
+  trimmed <- datawizard::data_filter(x, Category != "Total")
+  trimmed$Category <- factor(trimmed$Category, levels = trimmed$Category)


When converting Category to a factor, it is safer to use unique(trimmed$Category) for the levels argument. While duplicates might not be expected in a single group means table, using unique() is a more robust and defensive programming practice.

trimmed$Category <- factor(trimmed$Category, levels = unique(trimmed$Category))

These can never be duplicates because of how the object is created.

gemini-code-assist · 2026-07-06T20:44:33Z

+
+  x_attributes <- lapply(x, attributes)
+  x_var_names <- lapply(x_attributes, `[[`, "var_mean_label")
+  x_captions <- mapply(.build_caption, x, x_var_names)


Using mapply with list arguments that can contain NULL elements (such as x_var_names if the "var_mean_label" attribute is missing) can lead to unexpected recycling behavior or errors in R. Using vapply with an index is much safer and guarantees a character vector output.

x_captions <- vapply(seq_along(x), function(i) {\n .build_caption(x[[i]], x_var_names[[i]])\n }, character(1))

gemini-code-assist · 2026-07-06T20:44:33Z

+  p <- p + plotlist
+  print(p)


S3 plot methods should return the plot object and let the auto-printing mechanism or the user handle printing. Explicitly calling print(p) inside the function causes the plot to be drawn twice in interactive sessions when assigned to a variable (e.g., p <- plot(x)).

p + plotlist

This is incorrect when the plot is produced inside a loop.

gemini-code-assist · 2026-07-06T20:44:33Z

+  if (is.null(proportion_type)) {
+    x_long
+    plotlist[[1]] <- aes(x = row_var, y = name, fill = value)


The standalone expression x_long on line 16 is redundant and does nothing. It should be removed.

if (is.null(proportion_type)) {\n plotlist[[1]] <- aes(x = row_var, y = name, fill = value)

Not supposed to be part of PR

bwiernik · 2026-07-06T22:29:49Z

There are a variety of places where this code won't work as expected. Could you give a reproducible example of a case where the levels of a factor aren't being preserved as expected with the current version of the package?

Note--if you haven't converted the variable to factor before modeling, that is the best place to do so

strengejacke · 2026-07-07T06:22:18Z

We already have methods for plotting objects returned by means_by_group(), you should build upon that code instead of re-implementing a second plot() method, especially, since plot() for that class (.dw_groupmeans) is already defined in the datawizard package.

The relevant code is here:
https://github.com/easystats/see/blob/main/R/plot.means_by_group.R

The current implementation is also more elegant as it handles arguments directly, which makes the code in parts shorter and easier to maintain. We don't need to capture all the arguments via ....

elinw · 2026-07-07T18:28:37Z

Sorry sent the wrong branch

elinw · 2026-07-07T18:56:33Z

Hi

We already have methods for plotting objects returned by means_by_group(), you should build upon that code instead of re-implementing a second plot() method, especially, since plot() for that class (.dw_groupmeans) is already defined in the datawizard package.

The relevant code is here: https://github.com/easystats/see/blob/main/R/plot.means_by_group.R

The current implementation is also more elegant as it handles arguments directly, which makes the code in parts shorter and easier to maintain. We don't need to capture all the arguments via ....

Lol that was contributed by me, though in fact ... methods are easier to maintain because you don't have to add everything users suggest and they don't violate the OOP principle of matching the signature of the parent method.

elinw and others added 10 commits June 28, 2026 00:11

Add functions for plotting dw_groupmeans objects.

0430609

Merge branch 'main' into plot_groupmeans

a507dc4

revise

f12e4b1

change class attribute

5c37932

Updates based on review

9b67494

Remove unneeded addition

revert some odd changes

2bc462e

add back files

3a6d456

Some missed items

ce89635

Fix up some details.

8bd7391

Preserve the order of the data frame when making the plot.

6a65f7b

gemini-code-assist Bot reviewed Jul 6, 2026

View reviewed changes

Merge branch 'main' into plot_groupmeans_plot

0af48a5

Merge branch 'main' into plot_groupmeans_plot

8b92e8b

elinw closed this Jul 7, 2026

elinw mentioned this pull request Jul 7, 2026

Keep category order in group means plot #450

Closed


		x_long <- do.call(rbind, x)

		trimmed <- datawizard::data_filter(x_long, Category != "Total")

Uh oh!

Uh oh!

Conversation

elinw commented Jul 6, 2026

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

elinw Jul 7, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

elinw Jul 7, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

elinw Jul 7, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

elinw Jul 7, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

elinw Jul 7, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

elinw Jul 7, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

elinw Jul 7, 2026

Choose a reason for hiding this comment

Uh oh!

bwiernik commented Jul 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

strengejacke commented Jul 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

elinw commented Jul 7, 2026

Uh oh!

elinw commented Jul 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

bwiernik commented Jul 6, 2026 •

edited

Loading

strengejacke commented Jul 7, 2026 •

edited

Loading