Time Fixed Effect | Year Fixed Effect | Country Fixed Effect in Stata Part 3

[@thedatahall]

The following link contains the files used in this video: payhip.com/b/iR9Tn

[@viveksarati]

Since industry and country is time invariant, does it mean it is best to use Random effect model instead of FE eventhough the hausman test is requesting for FE model?

[@ashevasaefudin]

Thanks A lot. I just need it for my thesis

[@ammaralghamri4901]

excellent explanation, excellent presentation. absolutely outstanding

[@thedatahall]

Thanks for appreciation

[@SurenAntonPerera]

Thanks so much for clear and precise guidance.

[@thedatahall]

Glad it helped

[@Vyquach0]

after concluding that we have time and entity fixed effect in the model, I test Autocorrelation and heteroscedasticity both exist in FEM, what should I do next, please replyyyyyyy

[@bhuyashitalukdar2685]

Sir at first thanks a lot for this wonderful lecture it was really helpful. I had doubts about my data. If you could clarify. The independent variable in my dataset is a time invariant dummy variable. When I perform brush pagan test, it suggests there exists a panel effect. Then when I perform hausman test it suggest to use FE mode. However that is not possible for my data. How do I proceed? Can I use RE model.

[@Ganieirfan]

Hello Sir, Thank you so much for your wonderful videos on regression using stata. You have made it so easy to understand panel data analysis. I have one question. I am having panel data where I have set comapny code and year as panel variables using xtset command. Now when I choose between random effects and fixed effects hausman test suggests fixed effects p

[@thedatahall]

Hi, thanks for the appreciation. As explained in part 1 or 2, we cannot use time invarient variables with fe option in xtreg. Because industry is not changing from one year to another therefore the xtreg with fe option will ommit the perfect collinear variables. In this case u should use reg y x i.industry i.firm.

[@Ganieirfan]

@@thedatahall Thank you for responding. I have more than 700 firms and for 22 years. if i use i.firm it doesn't show me results because of number of firms. Is it correct to use xtreg with i.industry and re option, even though hausman test doesn't suggest re.?

[@thedatahall]

Because there r lots of dummies it will take time to show the results. Or u can generate dummies manually for each firm and input them in the regression. I have discussed this in part 2. But it is not correct to use random effect

[@Ganieirfan]

@@thedatahall are you saying that if I use firm dummy it takes care of fixed effects which we use normally and putting industry dummy with it takes care of industry effects? (As simple 'reg" command doesn't take into consideration the panel structure of data. Can I use xtreg fe and use interaction between industry dummy and my independent variable, eg (i.ind*x) will that be correct as it will take care of omission due to multicolineraty using fe with xtreg and industry dummy. I am sorry to ask too many questions, may be stupid too.

[@thedatahall]

Have you watched all the 5 parts? In these videos i have explained that taking firm dummies with regress command or using xtreg with h fe option are same. In finance papers its common to use industry fixed effect but what they do is regress y x i.industry. there is no way to use xtreg,fe and also incorporate any time invarient variables. Even if u do xtreg y x,fe i.e. without industry fixed effect, that is fine too because the main idea of firm fixed effect was to adjust for time invarient variables, so it automatically takes care of that.