Estimation Workflows¶

TabDat provides a comprehensive suite of econometric, statistical, machine learning, and Bayesian estimation commands.

1. Linear & Quantile Regression¶

Classical, Robust & Clustered OLS¶

tabdat> regress wage educ exper
tabdat> regress wage educ exper, robust
tabdat> regress wage educ exper, cluster(industry)

Quantile Regression¶

Estimate median or arbitrary conditional quantiles:

tabdat> qreg wage educ exper, quantile(0.5)
tabdat> qreg wage educ exper, quantile(0.9)

2. Binary & Limited Dependent Variables¶

Logit & Probit MLE¶

tabdat> logit employed educ age married, robust
tabdat> probit employed educ age married

Censored Regression (Tobit)¶

tabdat> tobit hours_worked wage educ, ll(0)

Sample Selection (Heckman)¶

Two-step Heckman selection estimator:

tabdat> heckman wage educ, selectdep(inwork) select(educ age kids)

3. Count & Survival Data¶

Poisson MLE: poisson visits age insurance
Negative Binomial: nbreg visits age insurance
Zero-Inflated Poisson: zip visits age insurance, inflate(age)
Zero-Inflated Negative Binomial: zinb visits age insurance, inflate(age)
Parametric Survival: streg duration age, failure(event) dist(weibull)

4. Panel Data & Instrumental Variables¶

Panel Regression (`xtreg`)¶

Requires panel id_var time_var first:

tabdat> panel firm year
tabdat> xtreg investment capital profit, fe
tabdat> xtreg investment capital profit, re

Dynamic Panel GMM (`xtabond`)¶

Arellano-Bond linear dynamic panel estimator:

tabdat> xtabond investment capital profit, lags(1)

Instrumental Variables (`ivregress`)¶

Two-stage least squares (2SLS) and GMM:

tabdat> ivregress 2sls wage exper, endog(educ) iv(near_college parents_educ)

5. Causal Inference¶

Difference-in-Differences (`did`)¶

tabdat> did outcome, treat(treated) post(post_period)

Doubly Robust Difference-in-Differences (`drdid`)¶

tabdat> drdid outcome covariates, treat(treated) post(post_period)

6. Regularization & Machine Learning¶

Lasso: lasso linear y x1 x2 x3
Post-Lasso OLS: postlasso linear y x1 x2 x3
Ridge Regression: ridge linear y x1 x2 x3
Elastic Net: elasticnet linear y x1 x2 x3, l1_ratio(0.5)
Cross-Validated Models: cvlasso linear y x1 x2 x3, cvridge ..., cvelasticnet ...
Double/Debiased Machine Learning (DML): dml linear y controls, treat(t)

7. Bayesian Estimation¶

Bayesian Ridge Linear Regression¶

tabdat> bayes linear wage educ exper

Full MCMC Sampling (`bayes:` Prefix)¶

MCMC estimation via Bambi/PyMC backends with custom priors:

tabdat> bayes, draws(2000) burnin(1000) chains(4): regress wage educ exper

MCMC Diagnostics: estat bayes
MCMC Trace Plots: bayesplot trace, bayesplot density, bayesplot autocorrelation
Posterior Predictive Predictions: predict wage_pp, posterior_predictive std interval

8. Spatial Econometrics (`spregress`)¶

Fit spatial lag (SAR) and spatial error (SEM) models:

tabdat> spregress crime income, coord(lat lon) model(lag)

9. Post-Estimation Diagnostics¶

predict <varname>, xb: Compute linear predictions.
predict <varname>, residuals: Compute residuals.
predict <varname>, pr: Predicted probabilities (after Logit/Probit).
estat vif: Multicollinearity variance inflation factors.
estat report: Generate self-contained HTML regression summary report.
test x1 = x2: Linear hypothesis Wald tests.
lincom x1 + 2*x2: Estimate linear combinations of parameters.