Feature request: Add bin_prop computed variable to stat_bin for proportion-based histograms

[kieran-mace]

Summary

stat_bin currently lacks the after_stat(prop) functionality that stat_count provides, making it difficult to create proportion-based visualizations for continuous data. This feature request proposes adding a bin_prop computed variable to stat_bin to achieve feature parity.

Problem Description

Currently, users can create proportion-based bar charts with discrete data using stat_count:

# This works with discrete data
ggplot(data, aes(x = discrete_var, y = after_stat(prop), fill = group)) +
  geom_bar(position = "dodge")

However, there's no equivalent for continuous data with stat_bin:

# This doesn't work - no prop variable available
ggplot(data, aes(x = continuous_var, y = after_stat(prop), fill = group)) +
  geom_histogram(position = "dodge", bins = 10)

Use Case Example

Consider analyzing weight distribution by sex. Users want to see the proportion of each sex within weight bins:

# Desired functionality (currently not possible)
ggplot(people_data, aes(x = weight, y = after_stat(bin_prop), fill = sex)) +
  stat_bin(geom = "col", bins = 8, position = "dodge") +
  scale_y_continuous(labels = scales::percent) +
  labs(y = "Proportion within bin")

This would show insights such as:

• Lower weight bins: ~100% female
• Middle weight bins: Mixed proportions
• Higher weight bins: ~100% male

Something like this:

Proposed Solution

Add a bin_prop computed variable to stat_bin that calculates the proportion of each group within each bin:

• bin_prop = count_in_group / total_count_in_bin
• Handles multiple groups and respects weights
• For single groups: bin_prop = 1 (backwards compatible)
• For empty bins: bin_prop = 0

Benefits

1. Feature parity with stat_count
2. Enables proportion-based histograms for continuous data
3. Useful for demographic analysis and group comparisons
4. Backwards compatible - doesn't break existing code

Alternatives Considered

1. Manual calculation: Users could manually calculate proportions, but this is cumbersome and error-prone
2. Using stat_count with discretized data: Loses the benefits of proper binning algorithms
3. Custom stat function: Would require users to write their own implementation

Expected API

# Documentation would include:
#' @eval rd_computed_vars(
#'   count    = "number of points in bin.",
#'   density  = "density of points in bin, scaled to integrate to 1.",
#'   ncount   = "count, scaled to a maximum of 1.",
#'   ndensity = "density, scaled to a maximum of 1.",
#'   width    = "widths of bins.",
#'   bin_prop = "proportion of points in bin that belong to each group."
#' )

This would enable the intuitive usage:

aes(y = after_stat(bin_prop))

Additional Context

This feature would be particularly valuable for:

• Demographic analysis (age/income by group)
• Scientific data (measurements by treatment group)
• Market research (customer segments by behavior)
• Any scenario where you want to show group composition within continuous ranges

The implementation should handle edge cases like empty bins, single groups, and weighted data appropriately.

[kieran-mace]

PR: #6477