Skip to main content

Run Preliminary Econometrics Models

Econometrics helps users run preliminary models directly in Ecodata. The goal is quick screening before exporting data for deeper analysis.

12 Model Families, 40 Sub-models, and 105 Estimators Supported​

EcoData integrates a comprehensive Econometrics Engine allowing preliminary screening and estimation with:

12 Major Model Families​

  1. Classical Linear Regression: OLS, WLS, GLS, TLS (Total Least Squares).
  2. Regularized Regression: Ridge, Lasso, Elastic Net, Adaptive Lasso.
  3. Linear Panel Data: Pooled OLS, Fixed Effects (Entity/Time FE), Random Effects, Between Effects.
  4. Dynamic Panel Data: Arellano-Bond (Difference GMM), Blundell-Bond (System GMM).
  5. Limited Dependent Variable: Binary Logit, Binary Probit, Tobit (Censored), Truncated Regression, Heckman Selection (Heckit).
  6. Count Data: Poisson Regression, Negative Binomial, Zero-Inflated Poisson (ZIP), Zero-Inflated Negative Binomial (ZINB).
  7. Quantile Regression: Linear Quantile Regression, Panel Quantile Regression (FE-QR).
  8. Univariate Time Series: AR, MA, ARMA, ARIMA, SARIMA, ARCH, GARCH, EGARCH.
  9. Multivariate Time Series: Vector Autoregression (VAR), Vector Error Correction Model (VECM), Structural VAR (SVAR).
  10. IV & Simultaneous Equations: IV/2SLS, 3SLS, Seemingly Unrelated Regressions (SUR).
  11. Non-linear & Semi-parametric: Non-linear Least Squares (NLS), Generalized Additive Models (GAM).
  12. Causal Inference & Impact Evaluation: Difference-in-Differences (DiD), Propensity Score Matching (PSM), Regression Discontinuity Design (RDD).

Breakdown of 40 Sub-models and 105 Estimators​

The system provides 40 sub-models corresponding to different mathematical specifications. To ensure robustness of coefficients for academic research, the system supports 105 estimators configured based on the combination of:

  • Optimization Methods: OLS, FGLS, Maximum Likelihood (MLE), Quasi-MLE, GMM (1-step/2-step with Windmeijer correction).
  • Standard Error Covariance Structures: Homoskedastic, White Robust (HC0, HC1, HC2, HC3), Clustered Standard Errors (by Entity, Time, or Multi-way Clustering) to control for autocorrelation and heteroskedasticity.

Workflow​

  1. Open Econometrics.
  2. Choose a model category.
  3. Select the model.
  4. Select the dependent indicator.
  5. Select independent indicators from /api/indicators.
  6. Choose year range.
  7. Add custom variables if needed.
  8. Run the analysis.
  9. Review coefficients, standard errors, t-statistics, p-values, model fit, and interpretation.

Choosing Model Indicators​

ComponentExamples
Dependent variableGDP growth, poverty rate, export value, stock return.
Main explanatory variableFDI, PCI, tariff, education, investment.
ControlsPopulation, labor force, CPI, sector structure, year.
Time rangeChoose a period with enough observations for the model.

Citation Style​

The module supports citation styles such as APA 7, Chicago 17, Harvard 2008, IEEE 2008, and MLA 7. Keep indicator citations together with exported data and metadata.

Limits​

Econometrics is suitable for quick checks. For academic or official reports, export the data and re-test it in R, Python, Stata, or another specialized environment. For panel models, verify entity keys, time indexes, missing values, and aggregation before interpreting results.

Combining With AI Chat​

AI Chat can help choose a model, explain coefficients, remind users to check multicollinearity, missing data, stationarity, lag structure, or suggest alternate indicators when coverage is insufficient.