Econometrics Estimation & Modeling
EcoLab supports a comprehensive array of over 30 econometric models, handling various data structures such as cross-sectional, time-series, and panel data. The modeling module connects directly to powerful computation engines running Python, R (via the RMCP gateway), and Stata.
1. Supported Model Classification
The system categorizes models into 5 specialized groups:
| Model Group | Estimators | Application |
|---|---|---|
| Classical & Regularization | OLS (Ordinary Least Squares), WLS (Weighted), GLS (Generalized), Ridge, Lasso, Elastic Net | Baseline models, classical linear regression, and multicollinearity mitigation using regularization methods. |
| Panel Data | Fixed Effects (FE), Random Effects (RE), Between Estimator, GMM (Arellano-Bond, Blundell-Bond), IV/2SLS, 3SLS | Best for datasets tracking multiple entities (firms, countries) over several time periods. FE/RE helps control for unobserved time-invariant heterogeneity. |
| Time Series | ARIMA, GARCH, EGARCH, VAR (Vector AutoRegression) | Financial volatility modeling, macroeconomic business cycles, and forecasting. |
| Limited Dependent & Count | Logit, Probit, Tobit, Poisson Regression, Negative Binomial Regression | Used when the dependent variable is binary (0/1), censored, or a non-negative integer. |
| Causal Inference | PSM (Propensity Score Matching), DiD (Difference-in-Differences), RDD (Regression Discontinuity), IV/2SLS | Evaluates the impact of policy interventions or economic events while addressing endogeneity issues. |
2. Estimation Workflow
- In the Mô hình hóa (Modeling) module, click the Thêm mô hình (Add Model) button to open the configuration panel.
- Define Model Specification:
- Select the model group and estimator (e.g., Panel Data → Fixed Effects).
- Select the Dependent Variable ($Y$) from the parsed data columns.
- Select the Independent and Control Variables ($X_1, X_2, \dots, X_k$).
- Advanced Configurations (Optional):
- For panel data: Specify the Entity and Time identifier variables.
- For IV/2SLS: Select the endogenous variable(s) and their matching instrumental variables.
- Select standard error corrections: Homoskedastic, Robust, or Clustered (by entity/time).
- Click Chạy mô hình (Run Model) to send the request to the computation server.
3. Reviewing Estimation Results
Results are returned instantly and organized across 4 tabs:
Estimation Tab
Presents the regression coefficients table in academic format:
- Estimated coefficients ($\beta$), Standard Errors, $t$-stat (or $z$-stat) values, and $p$-values.
- Statistical significance is represented by standard asterisks:
***($p < 0.01$),**($p < 0.05$),*($p < 0.1$). - Model summary metrics: Coefficient of determination $R^2$, Adj-$R^2$, $F$-statistic, Log-Likelihood, AIC, and BIC.
Diagnostics Tab
Automatically runs standard statistical tests to verify classical regression assumptions:
- Breusch-Pagan / White Test: Tests for Heteroskedasticity (non-constant error variance).
- Durbin-Watson / Ljung-Box Test: Tests for Autocorrelation in errors.
- Jarque-Bera Test: Tests for Normality of the residuals.
- Hausman Test: (For panel data) Decides between Fixed Effects (FE) and Random Effects (RE).
Robustness Tab
Checks the sensitivity of your estimation:
- Runs regressions on subsamples (subsample analysis) to see if results change when removing outliers.
- Estimates alternative specifications (e.g., using alternative proxies or adding/removing controls).
Replication Code Tab
The system automatically compiles and displays the replication script in Python, R, or Stata. You can copy this code to run locally for verification and reporting.