Dear researchers, if panel data does not exhibit normal distribution, then fixed effects or random effects estimation technique would be suitable to apply or not?
to convert data to normally distributed,, we can transform data into log form.....but despite taking log of variables,,,,still distribution is not normal....then any other alternate method to make data distribution normal? comment plz
Assuming you have a non-normal multivariate distribution (conditional on exogenous variable), you can try Maximum Likelihood estimations, which should be asymptotically normal. There is no guarantee, however, that the standard errors will be true. Hence, you could be at the risk of generating false p-values and CIs. Or, you could get lucky.
You could also try QML (quasi-maximum likelihood), which is more accommodating of normality assumptions for standard errors and usually produces the same coefficient.
Check your software package's documentation.
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Biology
- Molecular Methods
- Cloning