Good afternoon,
I am doing my thesis on the impact of banking crises in education. My dependent variable is expressed as ratios (education) and my explanatory variable is the occurrence or not of a banking crisis (dummy variable). In addition, I also consider control variables (some are dummies others are specific values)
In this way, I would like to understand what is the best model to use. As I have ratios it makes more sense to consider the difference between my dependent variables right? And how do I distinguish my dummy variables from non-dummy variables in my model?
Basically, I would like to choose the fittest model for this situation.
I would really appreciate some help.
Thank you very much,
Carlota
You can run pooled OLS, and then can use 2SLS, Two Step System GMM for robustness check. For detail you can read this book as well:
https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=mostly+harmless+econometrics&oq=mostly#d=gs_qabs&u=%23p%3Daaa1AVN1VRUJ
Hi! You should choose between OLS, fixed and random effects based on the tests (Hausman and Breusch-Pagano tests). You can perform them in Stata or in any other software that you use.
If you have a ratio as a dependent variable, you can leave it as it is, you can take growth rate of the ratio or you can take a logarithm. In order to choose the best option, you should find a paper from a good journal where authors study the issue you are focusing on.
You can write me if you need more help with that.
Best regards,
Tatiana
there so many models you can use with panel data like, GMM, 2SLS, PTR, and PARDL
Or use the least square dummy variable technique for fixed effects.
ratio for dependent variable can remain. explanatory variables if in absolute figure then log transform but categorical factors can remain before regressing by panel estimation.. LSDV.
Perform unit root test e.g Johansen test, then do cointegration otherwise use Likelihood models
- Badges
- Science method
- Similar topics
- Teaching