Add support for non-numeric (categorical) predictors in discord_data()

[Copilot]

Overview

This PR adds support for non-numeric (categorical) predictors in the discord_data() and make_mean_diffs() functions, resolving the issue where predicting outcomes based on categorical variables like location ("south" or "north") would fail.

Problem

Previously, the make_mean_diffs() function attempted to compute differences and means for all variables, including categorical ones. This caused errors or warnings when non-numeric predictors were used, as operations like subtraction and mean calculation are not meaningful for categorical data.

Solution

Modified both implementations of make_mean_diffs() to:

1. Detect variable types using is.numeric() before processing
2. For numeric variables: calculate differences and means as before
3. For non-numeric variables: preserve individual values in _1 and _2 columns, set _diff and _mean to NA

Example Usage

library(discord)

# Create data with categorical predictor
data <- data.frame(
  id = 1:10,
  age_s1 = c(25, 30, 35, 40, 45, 50, 55, 60, 65, 70),
  age_s2 = c(23, 28, 33, 38, 43, 48, 53, 58, 63, 68),
  location_s1 = c("south", "north", "south", "north", "south", 
                  "north", "south", "north", "south", "north"),
  location_s2 = c("south", "south", "north", "north", "south", 
                  "south", "north", "north", "south", "south")
)

# Process with categorical predictor
result <- discord_data(
  data = data,
  outcome = "age",
  predictors = "location",
  id = "id",
  sex = NULL,
  race = NULL,
  pair_identifiers = c("_s1", "_s2"),
  demographics = "none"
)

# Result contains:
# - location_1, location_2: categorical values preserved ("south"/"north")
# - location_diff, location_mean: NA (not meaningful for categorical)
# - age_diff, age_mean: properly calculated (numeric outcome)

Changes Made

Core Functions

• make_mean_diffs_ram_optimized(): Added type checking and conditional processing
• make_mean_diffs_fast(): Added same logic for vectorized operations
• Removed unnecessary suppressMessages() and suppressWarnings() since type handling is now explicit

Documentation

• Updated roxygen comments to explain categorical variable behavior
• Enhanced discord_data() parameter documentation with categorical predictor examples
• Added NEWS.md entry documenting the feature

Tests

• Created comprehensive test suite (test-categorical_predictors.R) covering:
  • Character string predictors
  • Mixed numeric and categorical predictors
  • Factor variables
  • Both fast=TRUE and fast=FALSE modes

Use Cases

This enhancement enables researchers to:

• Use demographic variables as predictors (region, city, occupation type)
• Include categorical characteristics that vary between siblings
• Analyze mixed effects with both continuous and categorical predictors
• Predict outcomes based on location, school attended, or other categorical factors

Backward Compatibility

✅ All changes are backward compatible:

• Existing numeric predictor functionality unchanged
• No breaking changes to the API
• Existing code continues to work without modification

Fixes issue about creating discordant kinship data with non-numeric predictors.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Create discordant kinship data with non-numeric predictors</issue_title>
<issue_description>For example, if we predict age based on location and the location variable is encoded as "south" or "north". Update make_mean_diffs() to work with that and return the discordant-style data.</issue_description>

Comments on the Issue (you are @copilot in this section)

Fixes #5

---

💬 Share your feedback on Copilot coding agent for the chance to win a $200 gift card! Click here to start the survey.

[codecov]

Codecov Report

❌ Patch coverage is 88.88889% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 91.09%. Comparing base (368461d) to head (52be958).

Files with missing lines	Patch %	Lines
R/helpers_regression.R	88.88%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #35      +/-   ##
==========================================
- Coverage   91.20%   91.09%   -0.11%     
==========================================
  Files           7        7              
  Lines         853      865      +12     
==========================================
+ Hits          778      788      +10     
- Misses         75       77       +2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.