pysa-pv 0.1.8

PISA-style weighted least squares with BRR-Fay and plausible values.

pysa

pysa provides PISA-style weighted least squares (WLS) with BRR-Fay replicate weights and plausible values.

It is designed for analyses aligned with standard OECD/PISA practices:

• full-sample student weights
• 80 BRR–Fay replicate weights
• plausible value combination using Rubin-style rules (sampling + imputation variance)

This package is intended for research and reproducible workflows using PISA-like datasets.

---

Features

• Weighted least squares (WLS) using full-sample weights
• BRR-Fay variance estimation using replicate weights
• Plausible value (PV) combination
  • Averages coefficients across PVs
  • Combines sampling + between-PV variance into final standard errors
• Flexible model specification:
  • numeric regressors (num_vars)
  • categorical regressors (cat_vars, one-hot encoded with drop_first=True)
  • fixed effects (fixed_effects, also one-hot encoded)
• Optional country filtering
• Clean results object: PysaResult

---

Installation

From PyPI

pip install pysa-pv