I am dealing with panel data, n=28 and t=7. All the panel models including POLS, FD, FE, RE generate very low R squared values less than .10, but between variation model generated a r-squared of .96. Can somebody advise me?
I have always tell friends and students that R-square is a measure of explanatory power, and not a model fit indicator. A lot of reasons might be responsible for low R- squared, like you have used n = 28 (too few data point). But that does not invalidate the results. You should note that R-square, even when small, can be significantly different from 0, indicating that your regression model has statistically significant explanatory power. Most scientific researches don’t even have up to 20% R-squared, you can check (Scorpus, web of science, etc). Like professor R. T. Brennan said, “we should be caught in a trap of referring to r-square values as high or low without knowing much more”.
Conduct a diagnostic analysis to ascertain your model fit and explanatory power. Use robust white heteroscedasticity-consistent (Eicker-White) standard errors and apply another model to justify the results of the main model.
Above all make sure your pre-analysis is done and correct.
Good luck!
I am a student but to my understanding, you would now focus on AIC or P-value and check the fitted model with the smallest AIC as the best model or use rejection rule of P-value less than 0.05. I guess panel data can be looked at as longitudinal data. you can fit the two or models as below.
fit1
Dear Osarumwense Osabuohien-Irabor , and Criscent Birungi Thank you very much for all your comments, really appreciate your insights
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation