Post-Processing Scripts

The scripts/ directory contains stand-alone analysis and visualization utilities that operate on the outputs of the main pipeline. These are not part of the installable package and are intended to be run directly with Python from the project root.

When to Run These Scripts

Run scripts after the main pipeline has produced result files (Excel exports, CSV predictions, JSON parameters). The scripts require outputs from kinopt.

Script Reference

curve_similarity.py

Purpose: Compute per-row discrete Fréchet distances between "Observed" and "Estimated" curves from kinopt_results.xlsx. Useful for ranking best/worst model fits.

When to run: After kinopt optimization.

Inputs: - Excel file(s) with sheets named "Observed" and "Estimated" (wide format: one row per entity, columns are time points)

Outputs: - CSV file with per-row Fréchet scores (lower = better fit)

Example:

python scripts/curve_similarity.py \
  --kinopt-xlsx data/kinopt_results.xlsx \
  --out-dir results_scripts

Interpretation: - Fréchet distance = minimum "leash length" to walk both curves in order - Lower is better; 0 means identical curves - Scale-dependent: only compare values within the same dataset/scale