econometrics

name	econometrics
description	Comprehensive econometrics reference covering OLS, IV, MLE, GMM, panel data, limited dependent variables, treatment evaluation, and advanced topics. Use when advising on empirical methodology, interpreting regression output, explaining identification strategies, or selecting appropriate estimators.

Econometrics Skill

A comprehensive reference for applied econometrics synthesized from Cameron & Trivedi's Microeconometrics, Wooldridge's Econometric Analysis of Cross Section and Panel Data, and Greene's Econometric Analysis appendices.

Skill Structure

This skill is organized into modular markdown files covering major econometric topics:

Foundations

• foundations/asymptotic-theory.md - Convergence concepts, LLN, CLT, delta method
• foundations/probability-distributions.md - Key distributions for econometrics
• foundations/matrix-algebra.md - Essential matrix results

Core Methods

• core_methods/ols-linear-regression.md - OLS estimation, inference, diagnostics
• core_methods/instrumental-variables.md - 2SLS, GMM, weak instruments
• core_methods/maximum-likelihood.md - MLE, QMLE, information matrix
• core_methods/gmm.md - Generalized method of moments
• core_methods/hypothesis-testing.md - Wald, LR, LM tests; specification tests

Panel Data

• panel_data/linear-panel-models.md - FE, RE, first differences, Hausman test
• panel_data/dynamic-panels.md - Arellano-Bond, system GMM
• panel_data/nonlinear-panels.md - Binary, count, censored panel models

Limited Dependent Variables

• limited_dependent/binary-models.md - Probit, logit, marginal effects
• limited_dependent/multinomial-models.md - MNL, nested logit, mixed logit
• limited_dependent/tobit-selection.md - Censoring, truncation, Heckman selection
• limited_dependent/count-models.md - Poisson, negative binomial, overdispersion

Treatment Evaluation

• treatment_evaluation/treatment-effects.md - ATE, ATT, LATE framework
• treatment_evaluation/matching-propensity.md - Propensity scores, matching methods
• treatment_evaluation/diff-in-diff.md - DID, event studies, parallel trends
• treatment_evaluation/regression-discontinuity.md - Sharp and fuzzy RDD

Advanced Topics

• advanced_topics/clustering-stratification.md - Cluster-robust inference, survey methods
• advanced_topics/measurement-error.md - Classical errors, attenuation bias
• advanced_topics/missing-data.md - MCAR, MAR, imputation methods
• advanced_topics/bootstrap.md - Nonparametric, parametric, wild bootstrap
• advanced_topics/quantile-regression.md - LAD, conditional quantiles

Usage Guidelines

When to Use This Skill

• Advising on econometric methodology for empirical research
• Helping interpret regression output and diagnostics
• Explaining identification strategies and their assumptions
• Debugging estimation problems or unexpected results
• Selecting appropriate estimators for specific data structures

Key Principles

1. Identification first: Before discussing estimation, clarify what parameter is identified and under what assumptions.

2. Match estimator to data structure:

   • Cross-section → OLS, IV, or limited dependent variable models
   • Panel data → FE, RE, or dynamic panel methods
   • Repeated cross-sections → Pseudo-panels, DID

3. Robust inference by default: Always use heteroskedasticity-robust SEs; cluster when observations are grouped.

4. Specification over sophistication: Simpler models with credible identification beat complex models with dubious assumptions.

Common Pitfalls to Flag

• Using RE when FE assumptions are more credible
• Clustering at the wrong level
• Ignoring weak instruments
• Misinterpreting marginal effects in nonlinear models
• Assuming parallel trends without evidence
• Using inappropriate standard errors for DID with few clusters

Quick Reference: Estimator Selection

Data Structure	Main Concern	Recommended Approach
Cross-section, continuous Y	Endogeneity	IV/2SLS
Cross-section, binary Y	--	Probit/Logit
Panel, time-invariant confounders	FE vs RE	Hausman test; prefer FE if in doubt
Panel, lagged dependent variable	Nickell bias	Arellano-Bond GMM
Treatment evaluation	Selection on observables	Matching/propensity score
Treatment evaluation	Selection on unobservables	IV, DID, or RDD
Clustered data	Correlated errors	Cluster-robust SEs