The create_IV function previously failed when categorical variables (e.g., Gender, Department, Region) were supplied as predictors because it only supported numeric variables. This limitation prevented analysts from including valuable categorical factors in their Information Value analysis.

This PR also updates the package to v1.10.0, and is submitted into CRAN.

Changes Made

This PR extends the IV analysis pipeline to fully support categorical variables:

Core Implementation

• Modified calculate_IV function to detect and handle categorical variables using is.character() || is.factor()
• Updated binning strategy for categorical variables to use the categories themselves as intervals instead of quantile-based binning
• Enhanced labeling to display actual category names rather than numeric ranges
• Expanded map_IV function to include categorical variables in automatic predictor selection

Before and After

Before:

# This would fail
create_IV(data, outcome = "high_engagement", predictors = c("Gender", "Department"))

After:

# Now works seamlessly
create_IV(data, outcome = "high_engagement", predictors = c("Gender", "Department"))
create_IV(data, outcome = "promotion_rate", predictors = c("Level", "Region", "Tenure_Years"))
create_IV(data, outcome = "retention", predictors = NULL)  # Auto-includes categoricals

Example Output

For a categorical variable like FunctionType, the function now produces meaningful results:

  FunctionType  N Percent       WOE        IV
1  Engineering 63    0.63  0.673255 0.2612197
2      G_and_A 21    0.21 -1.239064 0.5755636
3   Operations 16    0.16 -0.833599 0.6878052

Business Impact

This enhancement enables analysts to:

• Include department, role, location, and other categorical factors in IV analysis
• Discover patterns in categorical variables that were previously inaccessible
• Build more comprehensive predictive models combining both numeric and categorical predictors
• Identify high-value categorical predictors for business insights

Testing

Added comprehensive test cases covering:

• Categorical predictors only
• Mixed categorical and numeric predictors
• Automatic inclusion of categorical variables when predictors=NULL
• Verification that existing numeric functionality remains unchanged

Fixes #230.

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