[josimar jesus]

Olá @Joao Santos Silva,
Na estimação de efeito fixo usando o módulo xtqreg qual função é usada? sigma(.) é a função identidade e Z = X?
Desde já agradeço,

[Joao Santos Silva]

Dear josimar jesus,

Yes, in xtqreg, sigma is specified as a linear function with Z = X.

Best regards,

Joao

[Mohamed Mahjoub Elheddad]

Joao Santos Silva Dear Joao ,

Do you have the STATA code for xtqreg?


Best,
Mohamed

[Joao Santos Silva]

Dear Mohamed Mahjoub Elheddad,

I am not sure if I understand your question, but you can download the code from SCC.

Best wishes,

Joao

[Ch Man]

Hello Mr. Joao,
I am working on new quantile panel data command for fixed effects "xtqreg" but my model includes one dummy variable which is my variable of interest and it gets dropped.

As you mentioned earlier that there is a possibility to get the estimates of fixed effects following the paper, can please elaborate a bit more over it.

[Joao Santos Silva]

Dear Ch Man,

The variable will be dropped if it does not vary over time. The fixed effects can be computed using the expression at the top of page 12 (see the link to the paper in the help file); all the ingredients needed to compute that expression are saved by the command: e(q), e(b_location), and e(b_scale).

Best wishes,

Joao

[Ch Man]

Dear Joao, Thank you for the response but I cannot find any expression at the top of page 12 of the paper it just explains a Lemma over it. May you please elaborate a bit more over it. I have seen the illustrative application of the Paper and there the dummy variable "OILIM" 's coefficient is mentioned in Table 4. My study some what follows the same line and I am interested in that coefficient. I would be a great help if you can elaborate over it.

Regards,
Ch Man

[Joao Santos Silva]

Dear Ch Man

The paper I am referring to is this one here, please see the first expression on page 12, just before Theorem 3.

Best wishes,

Joao

[Mohammad Azeem Khan]

Dear Joao Santos Silva

after running the xtqreg command in my case, the following message appered,

WARNING: some fitted values of the scale function are negative.
what does it mean?
does it pose any problem with the reliability of the results?


Thanks
Sincerely.

[Joao Santos Silva]

Dear Mohammad Azeem Khan,

That may be just noise but it may indicate that your scale function is not correctly specified.

Best wishes,

Joao

[Mohammad Azeem Khan]

Thanks a lot.
By the way, does this model work in a situation when some of the explanatory variables are endogenous?
and also can the lagged value of the dependent variable be included among the regressors?

[Luca Tiberti]

Dear @Joao Santos Silva,

I am trying to use xtqreg. However, coefficients do not vary across quantiles. Note that I have around 3,000 observations @ 2 periods. When I use "qreg", parameters change across quantiles.

An alternative way I tried to follow is by centering my Y and Xs variables to remove firm fixed effects. But quantiles would be estimated on the centered Y, rather than on the original Y (which would be incorrect).

Thanks in advance,
Luca

[Joao Santos Silva]

Dear Luca Tiberti,

I find it very strange that the parameters do not vary across quantiles. Can you please send the data and code to my email address so that I can check what is going on? Anyway, the estimator is not suitable with such short panel. Perhaps you can try using "correlated random effects" with clustered standard errors?

Best wishes,

Joao

[Joao Santos Silva]

Dear Luca,

Thank you for sharing the data and the code. As I suspected, the problem is caused by the fact that you only have data for 2 periods; in such extreme case it is not possible to identify the slopes of the different quantiles. xtqreg is only suitable when the number of time periods is "large".

Best wishes,

Joao

[Mohammad Azeem Khan]

Dear @Joao Santos Silva

You did a great help to me. I got good results in my models. One more favor, I need from your side is that is there any command to get the plots of the quantile estimates of this model?


Thanks Sincerely