WhatIf implements methods from Gary King and Langche Zeng's 2006 and 2007 papers on the dangers of extreme counterfactuals. The software offers easy-to-apply tests to evaluate counterfactuals without requiring sensitivity testing over specified model classes. It was authored by Heather Stoll, Gary King, Langche Zeng, Christopher Gandrud, and Ben Sabath.
Use Cases
- Evaluate the reliability of causal inferences based on counterfactual scenarios.
- Detect model-dependence in statistical analyses when counterfactuals are far from observed data.
- Assess the sensitivity of substantive inferences to modeling choices not based on empirical evidence.
Strengths
- Methods are based on peer-reviewed publications in Political Analysis and International Studies Quarterly.
- Software implements specific tests to reveal the degree of model-dependence in counterfactual analysis.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Last update date is unknown; freshness unverified.
Provenance
- Source
- paperswithcode