I have a panel with data on federal states and ran PooledOLS, a White-Test, a Breusch-Pagan-Test and a Durbin-Watson-Test to show heteroskedacticity and positive autocorrelation. After that, I ran RE and FE Models followed by a Hausman-Test to show that FE (as expected) is more suitable.
What other options do I have to analyze and play around with the data set? Would be thankful for some brainstorming and input. Thanks!
For instance,
(1) Multilevel models, which allow you to estimate group-specific slopes.
(2) Mundlak approach.
(3) Haussman Taylor.
(4) Using IV with panel data or GMM.
You can show the sensitivity of your results by applying the dynamic OLS, fully modified OLS, and GMM techniques.
Dynamic Seemingly Unrelated Cointegrating (DSUR) estimator may help you to improve the quality of your results.
System GMM
Difference GMM
Quantile regression
Quantile-Quantile regression
GWR Model
2SLS
3SLS
Ridge Regression
Threshold regression
and many more...
fixed effect with driscoll kraay standard errors will cover for heteroscedasticity, auto-correlation and cross-sectional dependency.
Some of the methods include fixed effects models, random effects models, and pooled regression models.
Fixed effects models are used when there is no variation in the independent variables over time.
Random effects models are used when there is variation in the independent variables over time.
Pooled regression models are used when there is no variation in the dependent variable over time.
Dynamic panel data models, which allow for lagged dependent variables and other time-varying covariates, and panel data regression discontinuity designs, which allow for causal inference in non-experimental settings.
- Badges
- Science topic
- Similar topics
- Thesis Research