Bug: bcf() overrides user-supplied num_burnin to 10 if propensity_train is NULL

[ninjasoul]

Summary

When calling bcf() with a user-specified num_burnin (e.g., 1000) but leaving propensity_train = NULL (to have it estimated internally), the num_burnin parameter is overridden to 10.

Problem Details

This appears to be caused by a variable scope issue in R/bcf.R.

1. A user calls bcf(..., num_burnin = 1000, propensity_train = NULL).
2. The code enters the if (is.null(propensity_train)) block here.
3. Inside this block, the assignment num_burnin <- 10 is made here.
4. This assignment overwrites the num_burnin variable for the entire function scope, not just for the internal propensity model.
5. As a result, the main BCF sampling loop incorrectly uses num_burnin = 10 instead of the user's intended 1000.

Expected Behavior

The user-supplied num_burnin value should be used for the main BCF model's burn-in phase. The num_burnin <- 10 assignment was likely intended to be a local variable (e.g., propensity_burnin) used only for the internal GFR propensity score estimation.

Suggested Fix

Possibly rename the variable at R/bcf.R to a new, local variable (e.g., prop_num_burnin <- 10) and use that new variable for the GFR sampling within that if block.