Quantile regression

[Nicola Rossi]

Dear Stata users,
I would like to apply a regression model, which assesses the relationship between my independent and dependent variable, depending on the quartiles of a "third variable".
Initial linear model: Y=X1 + X2 + X3 + X4 + E.term. I would like to estimate this model at different levels (quartiles) of X3.
It should be a sort of quantile regression, but the quartiles are not set on the dependent variable as usual.

Any suggestions would be highly appreciated.
Best,
NR

[Joao Santos Silva]

Dear Nicola Rossi,

What you want to do is not quantile regression. As far as I understand, is a standard regression performed for different sub populations defined by X3.

Best wishes,

Joao

[Nicola Rossi]

Dear Professor Silva,

I would like to check how my main relationship varies across different conditions of X3. I was wondering if it was the case to use the command qreg. By the way, I agree, the most straightforward solution is just to run a standard regression with "by function" by X3's levels. I was also wondering if it makes sense to keep X3 among my independent variables or if it is better to remove X3 in regression models on subpopulations.

Thank you and Best wishes,
Nicola

[Frode Andre]

Dear Nicola,

I do not know what your Y and Xs are, but it sounds to me as if you aim to use X3 as a moderating variable? This is perfectly doable using your framework, and also by creating a moderating variable with a few subpopulations (determined by the quartiles). If you are in a position to say something more about the variables, perhaps others can contribute with more accurate suggestions.


Best, Frode

[Joao Santos Silva]

Dear Nicola Rossi,

I do not see a particular reason to use qreg in this context, but of course you can estimate quantile regressions for different subpopulations defined by X3. You can leave X3 in the model for each subpopulation and see if it has enough variation to produce meaningful results.

Best wishes,

Joao

[Nicola Rossi]

Thank you Frode Andre Joao Santos Silva for your prompt responses. I'll follow your suggestions.
Best,
Nicola