Econometric model & estimator catalog

EcoData/EcoLab integrates a comprehensive Econometrics Engine with 12 major model families, broken down into 40 sub-models and 105 estimators. This page is the overview map: it helps you choose the right family for your data structure and research question, and explains how the 105 estimators are formed.

Test feasibility before investing further

Run a quick estimation on real data to confirm your topic has data, statistical significance and reproducibility before writing a thesis or paper. Every estimation exports replication code for Stata/R/Python.

---

Map of the 12 model families

---

Family → sub-model table

#	Model family	Representative sub-models	When to use
1	Classical linear regression	OLS, WLS, GLS, TLS	Linear relationships, baseline cross-section
2	Regularized regression	Ridge, Lasso, Elastic Net, Adaptive Lasso	Many regressors, multicollinearity, variable selection
3	Linear panel data	Pooled OLS, Fixed Effects, Random Effects, Between	Many units × many periods
4	Dynamic panel data	Arellano-Bond (Diff GMM), Blundell-Bond (System GMM)	Lagged variable, endogeneity, large N small T
5	Limited dependent variable	Logit, Probit, Tobit, Truncated, Heckman	Binary, censored, sample-selected outcomes
6	Count data	Poisson, Negative Binomial, ZIP, ZINB	Count outcomes (non-negative integers)
7	Quantile regression	Linear Quantile, Panel FE-QR	Effects across different quantiles
8	Univariate time series	AR, MA, ARMA, ARIMA, SARIMA, ARCH, GARCH, EGARCH	Forecasting, volatility of one series
9	Multivariate time series	VAR, VECM, SVAR	Multi-variable systems, cointegration
10	IV & simultaneous equations	IV/2SLS, 3SLS, SUR	Endogeneity, equation systems
11	Non-linear & semi-parametric	NLS, GAM	Non-linear relationships
12	Causal inference	DiD, PSM, RDD	Policy impact evaluation

In addition, ARDL (Autoregressive Distributed Lag) supports long-run/short-run relationships for time series with mixed I(0)/I(1) integration orders.

---

How are the 105 estimators formed?

The 40 sub-models correspond to distinct mathematical specifications. To serve academic research that requires coefficient robustness, each sub-model can be combined with several optimization methods and standard-error structures — yielding 105 estimators.

Component	Options
Optimization method	OLS, FGLS, Maximum Likelihood (MLE), Quasi-MLE, GMM (1-step/2-step with Windmeijer correction)
Standard-error structure	Homoskedastic; White Robust (HC0, HC1, HC2, HC3); Clustered by Entity, Time or Multi-way

Robust standard errors

Choosing the right standard-error structure controls for heteroskedasticity and autocorrelation — a decisive factor for the reliability of statistical inference (t-stats, p-values, confidence intervals).

---

Estimation workflow

1. In the Modeling module, choose the family by data structure.
2. Choose the sub-model (specific specification).
3. Declare the dependent variable $Y$ and the independent variables $X_1, \dots, X_k$.
4. Choose the standard-error structure (Homoskedastic / Robust / Clustered) and advanced parameters.
5. Run and read the estimation table, diagnostics, robustness; export the replication code.

---

Choosing a model by data structure

Data structure	Preferred family
Cross-section, continuous $Y$	Classical linear regression; regularized if many regressors
Binary / discrete / censored $Y$	Limited dependent variable; count data
Panel (N units × T periods)	Linear panel (FE/RE); dynamic panel (GMM) if lagged variable
Single time series	ARIMA/SARIMA; ARCH/GARCH for volatility
Multiple time-series system	VAR/VECM/SVAR
Policy impact evaluation	DiD, PSM, RDD, IV

---

See also

• Estimation & Modeling — detailed workflow
• Family detail pages: Panel data (FEM/REM) · Dynamic panel (GMM) · ARDL · VECM · DiD
• Worked examples: FDI & growth (ARDL) · Public debt & growth (panel)