I have some specific questions: (1) If the Hausman test favors the RE model over the FE model, can I still proceed with using the FE model? (It is because in the management field, considering that FE is more prevalent in research papers.) Is the Hausman test an absolute criterion? (2) I am using a two-way model with i.time and i.industry. Can both FE and RE models be applied in this case, or is only FE suitable? (3) In one of your UA-cam videos, you mentioned that when time-invariant variables (e.g., gender) are included, the RE model [(cov(z_i, u_i) ≠ 0)] instead of FE model [(cov(z_i, u_i) = 0)] is more likely to be preferred. In my case, the independent variables consist of "diversity" measured by gender, age, and education level. As age is a time-variant variable, would it still be appropriate to favor the RE model? (4) The secondary panel data includes industry classifications with 2-digit and 3-digit numbers. When conducting research with industry as a factor, is there a preference for using 2-digit or 3-digit numbers? Or is it at the discretion of the researcher? (It is because there is limited specific explanation in previous studies). I have reached out to the authors, but they used different industry numbers in each case. Thank you. I am looking forward your response for my question. Sincerely, James
Isnt it possible that the denominator becomes negative? As the random effects estimator is not consistent it standard error will become larger than the one of fixed effects and hence the denominator can become positive?
Hi, I got a p-value of 0.26 on Hausman test. Does that mean that I must do the Random-effects model and reject the Fixed-effects? I am a little confused. Thank you!
I think the reason is because Chi square distribution is a right-skewed distribution ,therefore no W's value is under 0. On contrast, most of W's value is close to 0. When W's value is bigger than 0 enough, that will indicate this value is unlikely to get due to random.
Mr Lambert, you've probably heard it a lot, but I also want to say that I appreciate what you do! Thanks!
This is an incredible explanation. You've married the intuition with the formal definitions in a way that many cannot do.
Dear Professor, sending lots of love from Malaysia for your amazing lectures!
Thank you Ben, a huge shout out from a econometric student in China. 您的视频简洁明了，帮助了我很多，谢谢！！！
This video, particularly out of all your videos, was very clear and very easy to understand! Thank you very much for this!
Thanks a lot for the explanation, it really helped my out grasping the notion of model specification
Thanks a ton!! Your videos are short and crisp. You are an excellent teacher. Thanks again!!
very clear explanation! thank you very much!
You should definitely do more video series in econometrics!
Thank you very much for your instructions, Ben
thank you so much this was unbelievably helpful!!
Can you please briefly explain the difference between the Hausman test and the Durbin-Wu-Hausman test?
I have some specific questions:
(1) If the Hausman test favors the RE model over the FE model, can I still proceed with using the FE model? (It is because in the management field, considering that FE is more prevalent in research papers.) Is the Hausman test an absolute criterion?
(2) I am using a two-way model with i.time and i.industry. Can both FE and RE models be applied in this case, or is only FE suitable?
(3) In one of your UA-cam videos, you mentioned that when time-invariant variables (e.g., gender) are included, the RE model [(cov(z_i, u_i) ≠ 0)] instead of FE model [(cov(z_i, u_i) = 0)] is more likely to be preferred. In my case, the independent variables consist of "diversity" measured by gender, age, and education level. As age is a time-variant variable, would it still be appropriate to favor the RE model?
(4) The secondary panel data includes industry classifications with 2-digit and 3-digit numbers. When conducting research with industry as a factor, is there a preference for using 2-digit or 3-digit numbers? Or is it at the discretion of the researcher? (It is because there is limited specific explanation in previous studies). I have reached out to the authors, but they used different industry numbers in each case.
Thank you.
I am looking forward your response for my question.
Sincerely,
James
Super helpful video! Thank you very much!
absolutely great video. keep it up!
Isnt it possible that the denominator becomes negative? As the random effects estimator is not consistent it standard error will become larger than the one of fixed effects and hence the denominator can become positive?
If you already know that the RE is not consistent, then there wouldn't be any point in testing RE vs. FE.
Thanks and very helpful indeed
Hi, I got a p-value of 0.26 on Hausman test. Does that mean that I must do the Random-effects model and reject the Fixed-effects? I am a little confused. Thank you!
If you obtain a p-value of 0.26, you cannot reject H0, i.e. you can use either RE or FE, but you should use RE because it is more efficient.
thanks this video is quite helpful
you are a life saver
Thanks Ben.
awesome video
What are the names of the denominator from eviews? Is it s.e square of reg. Or standard error square. Plz someone tell me
I think it should be the square of standard error
I got a question, how come under H_1, the denominator of the Hausman statistic W is not negative?
Willy Chen i think it indicates that there is higher correlation when W is large.
I think the reason is because Chi square distribution is a right-skewed distribution ,therefore no W's value is under 0. On contrast, most of W's value is close to 0. When W's value is bigger than 0 enough, that will indicate this value is unlikely to get due to random.
Many thanks to you
THKS A LOT GRAND PROF...
I take the summary to Rstudio😋
thanks alot
🎉
big up
very poor explanation.
Thank you very much ! I 've learned a lot!