These results of testing panel data of 540banks in 4 countries. since i am not specialist in statistics, are these results logic? R square in both RE and FE tests confused me?
is that acceptable to have (unbalanced data, low R square, zero overall Rsquare)?
Hi Somya Mohamed ,
R-square is a measure of explanatory power, and a low value can be attributed to several factors/reasons. If the value is low and statistically significant, then your model can be deemed acceptable – provided key assumptions (statistical and theoretical) relevant to your model are not violated - thus a low R-square value is acceptable.
A balanced panel is one in which the number of time periods T is the same for all individuals I; otherwise the dataset is an unbalanced panel. Most papers restrict themselves to balanced panels, despite the fact that unbalanced panels are a norm. In the case of randomly missing data, most commands of softwares like Stata can be applied to unbalanced panels without causing inconsistency of the estimators.
If you are using Stata, you can use a special procedure for unbalanced panels to check for information regarding the extent to which the panel is unbalanced. For instance, after you “xtset” the panel, you can use the “xtdescribe” command to assess the magnitude to which the panel is unbalanced and take necessary steps from there. A zero overall R-square is an extreme scenario which requires one to revisit the specified and estimated model.
All the best!
Somya Mohamed ,
Though your Hausman test results points to the preference for the FE model, the worrying area of interest to relook into is the zero overall R-square - my view. I would suggest that check if your coefficient estimates are consistent with theory, relevant statistical conditions relating to the properties of your individual variables' observations, the extent to which panel is unbalanced, etc.
Regards - Elvis
Hausman test, p-value, coefficient, and SE are the essential elements you need to interpret. However, you may have your R-square on the table, but all I have mentioned are what you explain.
- Badges
- Science topic
- Similar topics
- Counseling Psychology
- Confusion