Add run_sim_with_uncertainty() — PsN vpc uncertainty sampling wrapper

[roninsightrx]

Summary

• Adds run_sim_with_uncertainty() which wraps PsN's vpc --uncertainty_sample=N to simulate from a NONMEM model while incorporating parameter uncertainty from the covariance step
• Samples N parameter vectors from the multivariate normal distribution defined by the covariance matrix, runs one NONMEM simulation per draw, and returns a single data.frame with a sample_id column
• Adds two private helpers: extract_fit_folder() (resolves run folder from a fit object) and read_vpc_sim_tables() (parses PsN vpc output directory structure)
• Bumps version to 0.0.0.9029

Usage

# Basic usage — pass a fit object from run_nlme() (must have been run with $COVARIANCE)
out <- run_sim_with_uncertainty(fit = my_fit, n_samples = 200)

# With a custom simulation dataset (different obs times / subjects)
out <- run_sim_with_uncertainty(fit_folder = "run1/", data = sim_dat, n_samples = 200)

Test plan

• 26 tests in test-run_sim_with_uncertainty.R
• 20 pass without NONMEM (pure R / file I/O)
• 6 skip unless Pharmpy is configured; mock call_psn() to test end-to-end pipeline without actually running PsN
• Verify FAIL 0 | WARN 0 with devtools::test(filter = "run_sim_with_uncertainty")

🤖 Generated with Claude Code