Synthetic Control Donor Selection Using Expected-Value Spillover Tests
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Synthetic control models face challenges in accurately estimating treatment effects due to spillover effects from interventions, where donor units are impacted by the intervention, leading to potential bias and unreliable causal effect determinations.
Innovation Solution
A method for selecting training donors by comparing expected post-intervention values to actual values, using time-series data before and after the intervention, to identify donors not affected by spillover effects, and training a synthetic control model on these donors, with sensitivity analysis to verify the causal effect's trustworthiness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If donor units are selected based on similarity to target unit, then the synthetic control model can be trained with relevant data, but spillover effects from interventions may impact the donor units, introducing bias
Solution Approach 1:
The patent applies preliminary action by comparing pre-intervention expected values with actual post-intervention values before finalizing donor selection. This advance comparison identifies and excludes donor units that experienced spillover effects, ensuring only independent donors are used in the synthetic control model, thus preventing bias before it can affect the causal effect estimation
Solution Approach 2:
The patent implements feedback by using the comparison results between expected and actual values to iteratively refine donor selection. The process continuously evaluates whether selected donors maintain independence from intervention effects, and adjusts the donor set accordingly, creating a closed-loop system that ensures both similarity and independence are maintained
2Quantity of substance
If all available donor units are used for training, then more data is available for model training, but including donors affected by spillover effects increases potential bias in the synthetic control model
Solution Approach 1:
The patent applies the taking out principle by extracting and removing donor units that show evidence of spillover effects from the available donor pool. By comparing expected versus actual values, the method identifies and separates contaminated donors from clean donors, using only the extracted clean subset for training the synthetic control model, thus eliminating bias while preserving data quality
Solution Approach 2:
The patent applies local quality by applying different selection criteria to different donor units based on their individual responses to the intervention. Rather than uniformly including or excluding all donors, the method evaluates each donor's local characteristics (whether it experienced spillover effects) and selects only those with the appropriate local quality (independence from intervention effects) for inclusion in the training set
Data Source
AI summary
Systems and methods for donor selection for synthetic control models are provided. When selecting donors for synthetic control models, it is important that the selected donors are not impacted by an intervention. To determine whether potential donors are impacted by the intervention, expected post-intervention values for each donor are determined based on data from before the intervention. The expected values are compared against actual values, and training donors are selected based on the comparisons. A synthetic control model can be trained using the selected training donors.