PyCrossTRF

PyCrossTRF implements cross‑sectional transfer‑function (TRF) and covariate‑augmented transfer‑function (CTRF) models for analyzing how outcomes (e.g., mortality, yields) respond to temperature across many regions. The package follows the approach of Chang et al. (2016) and is designed for applied economists working with panel climate–impact data.

Key Features

• TRF/CTRF estimation – CTRF builds polynomial–Fourier bases and fits OLS to recover the response curve and the minimum mortality temperature (MMT)
• Covariate handling – covariates are centered/standardized and interact with temperature to produce CTRFs
• Sieve bootstrap – SieveBootstrap generates residual‑based confidence bands via AR‑sieve resampling
• Visualization utilities – plotting.Plot overlays TRFs/CTRFs and diagnostics for publication‑quality figures

Quick Start

1. Prepare data

Your DataFrame must include:

Column	Description
dep	dependent variable (e.g., log mortality)
temp_q	temperature regressor
date	time index
fips	cross‑sectional identifier
other covariates	optional (e.g., trends, demographics)

2. Fit a TRF

import PyCrossTRF as ctrf

pq = {'p': 4, 'q': 2}  # polynomial/Fourier orders
model = ctrf.CTRF(df=df,
                  dep='ln_mortality',
                  temp_r='temp_q',
                  covariates=[],
                  pq_order=pq,
                  scale={'temp_q': 'raw'})   # normalization rule
model.estimate_trf()
ctrf.plotting.Plot(model).trf_plot()

3. Add Covariates (CTRF)

model = ctrf.CTRF(df=df,
                  dep='ln_mortality',
                  temp_r='temp_q',
                  covariates=['time_trend'],
                  pq_order=pq,
                  scale={'temp_q': 'raw', 'time_trend': 'linear'})
model.estimate_ctrf()
ctrf.plotting.Plot(model).ctrf_plot()

4. Bootstrap Confidence Bands

bs  = ctrf.SieveBootstrap(model, maxL=24, B=999)
boot = bs.run()

lb = ctrf.CTRF(df=boot, dep='y_star_tau025', temp_r='temp_q',
               covariates=[], pq_order=pq, scale={'temp_q':'raw'})
lb.estimate_trf()

ub = ctrf.CTRF(df=boot, dep='y_star_tau975', temp_r='temp_q',
               covariates=[], pq_order=pq, scale={'temp_q':'raw'})
ub.estimate_trf()

ctrf.plotting.Plot(ub).trf_plot()  # overlay bounds

Interpreting Results

• TRF curve – shows expected outcome for each normalized temperature quantile.
• MMT – the temperature quantile minimizing the outcome is calculated automatically.
• CTRF curves – shifts in the response curve under different covariate values illustrate heterogeneous effects.
• Bootstrap bands – 2.5%/97.5% envelopes for policy‑relevant uncertainty assessment.

Citation Suggestion

Lee, Jaeseok Sean. 2025. “PyCrossTRF: Semi-parametric cTRF estimation model”. https://github.com/SeanJSLee/PyCrossTRF.

Reference

Chang, Yoosoon, Chang Sik Kim, J. Isaac Miller, Joon Y. Park, and Sungkeun Park. 2016. "A new approach to modeling the effects of temperature fluctuations on monthly electricity demand." Energy Economics 60:206-216

License

MIT © 2024 Jaeseok Sean Lee