Excellent video. Anyway, why do you use , fe i(panel) option? can we use fe (i(time) as well? How do we interpret then? Can you suggest any further reading for me on this?
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
If the Hausman statistic is significant (p-value is low, typically < 0.05), it indicates that the fixed effects model is preferred, suggesting the presence of endogeneity. If the Hausman statistic is not significant, it implies that the random effects model is appropriate, and there is no evidence of endogeneity.
What doet is mean If positive coefficient (+b1) of independent variable (x1) in OLS regression change to negative one (-b1) in LSDV regression? How can it be interpreted?
My teacher told me that you can use the Breusch and Pagan Lagrangian multiplier test for random effects ( to see if " c " exist ) its command is : " xttest0 " ( I use this command right after I use the regress command - I don't know if it's necessary but I got used to it, so ... ) look at the P-value of the test ( It's at the end, where it say " Prob > chibar2 = ... " ) If the P-value is small ( for example , < 1% ) then you can say that " I'm 99% sure that " c " exist" " c " exist also mean you shouldn't choose P-OLS but RE or FE instead And I was told that POLS bring much less value/information than RE or FE since it consider our data as separated pieces so you will not use it regularly To choose between FE and RE, use Hausman test ( you can you the Help -> Search on Stata to find it, it's not very complicated. You just have to store your models' result by " estimate store " and remember to put FE before RE on the test ( sorry for my English, hope this comment helped )
for hausman test always use consistent estimation (fe) first and then efficient (re), you run it hausman re fe which is not correct?
Dk you have tutorials on correction of heteroscedasticity, autocorrelation, cross sectional dependence?
Excellent video. Anyway, why do you use , fe i(panel) option? can we use fe (i(time) as well? How do we interpret then? Can you suggest any further reading for me on this?
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
thanks for useful video, would you prepare and upload GMM and thanks alot
InSha Allah I will do it.
AOA sir! actually my Hausman test always fail and I don't understand how this happens
If the Hausman statistic is significant (p-value is low, typically < 0.05), it indicates that the fixed effects model is preferred, suggesting the presence of endogeneity.
If the Hausman statistic is not significant, it implies that the random effects model is appropriate, and there is no evidence of endogeneity.
What doet is mean If positive coefficient (+b1) of independent variable (x1) in OLS regression change to negative one (-b1) in LSDV regression? How can it be interpreted?
Can you provide this data for practice
Good explanation. I need command for chow test in Stata. Any one can help me please
Sir please make a video .. how to choose between OLS model and fixed / random Effect ...what are the test?
My teacher told me that you can use the Breusch and Pagan Lagrangian multiplier test for random effects ( to see if " c " exist )
its command is :
" xttest0 "
( I use this command right after I use the regress command - I don't know if it's necessary but I got used to it, so ... )
look at the P-value of the test ( It's at the end, where it say " Prob > chibar2 = ... " )
If the P-value is small ( for example , < 1% ) then you can say that " I'm 99% sure that " c " exist"
" c " exist also mean you shouldn't choose P-OLS but RE or FE instead
And I was told that POLS bring much less value/information than RE or FE since it consider our data as separated pieces so you will not use it regularly
To choose between FE and RE, use Hausman test ( you can you the Help -> Search on Stata to find it, it's not very complicated. You just have to store your models' result by " estimate store " and remember to put FE before RE on the test
( sorry for my English, hope this comment helped )
@@TungNguyen-hc6mithanks for your response. But when I ran BP test, and my p value is greater than 0.5? Then what to do?
Hi if F test pavlue is higher than 0.05 does it mean Pool ols is better than fixed effect
Nice content, I also have rural development and planning volume 1-3 on my UA-cam channel
Unfortunately, I have to inform anyone watching that this video is wrong in almost all aspects. Skip this one
what do you mean?