Hi, I conducted a Hausman test whereby I got the results as show in the screenshot. Both ROE and ROA shows fixed effects, and Tobin Q comes up with random effects. Which one should I use in this case? Or am I meant to run a fixed effects regression separately and then a random effects regression for Tobin Q?
Also, what is the difference between using Least Squares Dummy Variable (LSDV) approach vs Fixed effects panel data regression & Random effects panel data regression
Using LSDV, is the same as fixed effect, but under LSDV you can explore which firms/country/ individuals have high effect compared to the rest of the group. This is only the difference.
But remember, using fixed and random effects has drawback. It requires the variables to be strictly exogeneity. In real world, the variables are endogeneous.
So, I would strongly advise you to use GMM, if you have secondary data. If not, at least two stage least square (2SLS) would be okay.
Mwoya Byaro Thank you for the quick reply, would there still be a need to conduct a Hausman test if I were to use the LSDV method?
Additionally, if I were to stick with the fixed/random effect approach as opposed to the GMM/2SLS you suggested. From the results above, how would I justify using either of the models as I got mixed results?
if you have any steps to use fixed and random effect in the meta-analysis, please help me
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation