postshock: An R Package for Donor-Adjusted Forecasting After Structural Shocks
summary
The gist
We present the postshock R package, which implements and extends a donor-based framework for forecasting when a structural shock is known and the target response of interest is not yet observed This…
In short
The postshock package provides an R workflow to adjust forecasts for a target when a structural shock occurs before the target's response is observed. It estimates shock effects from past donor episodes, aggregates these effects using donor weights, and adds this adjustment to a baseline forecast. This method borrows information from completed outside events to estimate the mean or covariate-dependent mean of the shock effect.
Key concepts
- Donor-based framework
- This is a method where shock effects are estimated from historical 'donor' episodes. These estimates are then aggregated using weights to adjust a baseline forecast for a new, shocked target. The key is transferring estimated effects from completed outside episodes to inform the current forecast.
- Postshock adjustment
- This process involves adjusting an existing baseline forecast after a structural shock is known but the actual response of the target hasn't been measured yet. It uses aggregated donor-specific shock effects to modify the initial prediction, targeting a specific mean or covariate-dependent mean.
- Donor balancing mechanism (dbw)
- This function constructs weights for different donors so that their combination closely resembles the target in selected matching features. The goal is to find weights that make the weighted donor feature vector similar to the target's feature vector, controlled by a regularization parameter lambda.
- Conditional forecasting
- The package offers tools for generating forecasts based on specific conditions, such as conditional-mean forecasting (SynthPrediction) or conditional-variance forecasting (SynthVolForecast). This allows users to predict outcomes given certain known information or scenarios.
Terminology used across episodes
This episode discusses
The paper
postshock: An R Package for Donor-Adjusted Forecasting After Structural Shocks · Read on arXiv
University of Illinois Urbana-Champaign
Transcript
Introduction to the show: ident: Statistics. The day in statistics: methods, theory and the inference under them.
Tom: I'm Tom.
Jane: And I'm Jane, here with Lu, Meng and our in-house large language model, Lalam.
Tom: We're looking at the paper: "postshock: An R Package for Donor-Adjusted Forecasting After Structural Shocks".
Jane: Here's the gist of the paper. We present the postshock R package, which implements and extends a donor-based framework for forecasting when a structural shock is known and the target response of interest is not yet observed This package provides an end-to-end workflow for donor-effect estimation, donor weighting, target baseline forecasting, and post-shock adjustment.
Tom: The core methodological concept The central idea of Lin and Eck (2021) is to estimate shock effects from completed historical donor episodes, aggregate these estimated donor-specific effects, and use the resulting quantity to adjust a baseline forecast for the newly shocked target.
Jane: The distinctive feature of this framework is therefore not aggregation alone, but the information being transferred: estimated shock effects from completed outside episodes are borrowed before the corresponding response of the target is available.
Tom: This distinguishes donor-based adjustment from methods that aggregate complete forecasts or weight donor outcomes to construct counterfactual trajectories.
Paper summary: Tom: Moving into the summary of "postshock: An R Package for Donor-Adjusted Forecasting After Structural Shocks," we see they present an R package designed to forecast when a structural shock is known but you haven't observed the target response yet.
Jane: The main claim is that this package implements and extends a donor-based framework. It handles estimating shock effects from historical donor episodes, balances donors using specified matching features, and then transfers that adjustment into a baseline forecast for your target series.
Lu: The central idea they are building on is from Lin and Eck, two thousand twenty-one <ref:2610.11841#pg2>. They estimate shock effects from completed historical donor episodes, aggregate those estimated donor-specific effects, and use that resulting quantity to adjust a baseline forecast for the newly shocked target.
Meng: They distinguish this approach by saying it’s not just about aggregation; it's about transferring the information itself—borrowing estimated shock effects from completed outside episodes before you even see your target response.
Lalam: This distinguishes donor-based adjustment from methods that just aggregate complete forecasts or weight donor outcomes to build counterfactual trajectories. The aggregate adjustment targets the mean, or a covariate-dependent mean, of the shock-effect distribution rather than exact recovery of the realized effect.
Tom: Architecturally, they organize this as a modular workflow separating donor-effect estimation and weighting from your target-specific forecasting model. This allows for flexibility across different modeling approaches.
Jane: They provide interfaces like SynthPrediction for conditional mean forecasting through ARIMA and ARIMAX models, and SynthVolForecast for conditional variance forecasting through GARCH X models.
Lu: They also include extra functionality like structured donor pools, control-shock regressors, automated GARCH X order selection, and reproducible empirical workflows.
Meng: From a production standpoint, the focus on modularity is good because it means we can reuse the same weighting mechanism across different forecasting tasks if that works well for our specific needs.
Lalam: The paper is available as an R package, offering integrated workflows for conditional mean and variance forecasting after structural shocks.
Conclusion: Tom: Thinking about the title "postshock: An R Package for Donor-Adjusted Forecasting After Structural Shocks," it really boils down to providing a ready-made tool for this specific forecasting problem.
Jane: It’s essentially a way to systematically use historical donor episode data to adjust future forecasts when you know something big has changed but the outcome you care about is still in the future.
Lu: The authors are providing an end-to-end workflow that handles everything from estimating those donor effects to the final post-shock adjustment, with a modular design for flexibility.
Meng: For someone trying to deploy this, it means you get these pre-built interfaces for conditional mean and variance forecasts that you can plug into your existing system if it fits your modeling needs.
Lalam: The implication is that this framework offers a way to borrow information across related time series by balancing separate estimations of heterogeneous units against pooling information, which is important in panel data forecasting.
Tom: So the big picture here is using completed donor episodes to reduce forecast risk when structural shocks are present, as long as you accept targeting the mean rather than trying to hit the exact realized effect.
Jane: That’s a concrete summary: it's about reducing risk by borrowing information systematically from past donor behavior into future forecasts after a known structural change.
More episodes
- Daily Summary for 2026-11-19
- 2310.07973-Statistical Performance Guarantee for Subgroup Identification with Generic Machine Learning
- 2408.09155-Learning Robust Treatment Rules for Censored Data
- Daily Summary for 2026-11-19
- 2603.23397-Kinetic Langevin Splitting Schemes for Constrained Sampling
- 2606.09906-An information-geometric framework for mapping maximum potential biodiversity
- 2603.27718-Induced replication and the assessment of models
- 1307.0898-Entropy of a Zipfian Distributed Lexicon
- 2406.19974-Generalizing self-normalized importance sampling with couplings
- 2412.10213-Collaborative Optimal Design of Controlled Experiments with Subject Covariates