P Value Computation in Quantile Regression function of Statsmodel

47 views
Skip to first unread message

Him

unread,
Nov 4, 2022, 10:06:23 AM11/4/22
to pystatsmodels

Hi,

 

Can you please provide any resource where I can find the steps of computing the p-value for checking the significance of explanatory variable in the quantile regression function of statsmodel?

 

 

josef...@gmail.com

unread,
Nov 4, 2022, 10:10:25 AM11/4/22
to pystat...@googlegroups.com
quick answer:

AFAIR, it's described in the Stata docs, page 27 of

Josef

On Fri, Nov 4, 2022 at 10:06 AM Him <himadri...@gmail.com> wrote:

Hi,

 

Can you please provide any resource where I can find the steps of computing the p-value for checking the significance of explanatory variable in the quantile regression function of statsmodel?

 

 

--
You received this message because you are subscribed to the Google Groups "pystatsmodels" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pystatsmodel...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pystatsmodels/1f73faaf-2d9d-41de-b76d-015cdb90b67fn%40googlegroups.com.

josef...@gmail.com

unread,
Nov 4, 2022, 10:19:18 AM11/4/22
to pystat...@googlegroups.com
On Fri, Nov 4, 2022 at 10:10 AM <josef...@gmail.com> wrote:
quick answer:

AFAIR, it's described in the Stata docs, page 27 of


The docstring says

    The asymptotic covariance matrix is estimated following the procedure in
    Greene (2008, p.407-408), using either the logistic or gaussian kernels
    (kernel argument of the fit method).

but it's based on the approach in more recent version of Stata.
for example, the default kernel is Hall-Sheather, AFAICS/AFAICR.

josef...@gmail.com

unread,
Nov 4, 2022, 10:23:53 AM11/4/22
to pystat...@googlegroups.com
On Fri, Nov 4, 2022 at 10:19 AM <josef...@gmail.com> wrote:


On Fri, Nov 4, 2022 at 10:10 AM <josef...@gmail.com> wrote:
quick answer:

AFAIR, it's described in the Stata docs, page 27 of


The docstring says

    The asymptotic covariance matrix is estimated following the procedure in
    Greene (2008, p.407-408), using either the logistic or gaussian kernels
    (kernel argument of the fit method).

but it's based on the approach in more recent version of Stata.
for example, the default kernel is Hall-Sheather, AFAICS/AFAICR.


Also, QuantileRegression is an M-estimator similar to RLM with a sandwich cov_params.
However, I realized this only after the merge, so it ended up as a subclass of Regression models and not as a class similar to RLM.

Theory for cov_params is similar to median regression in the robust literature. V-shaped objective function with kink.

Josef 

Himadri Manna

unread,
Dec 7, 2022, 2:12:38 AM12/7/22
to pystatsmodels
SO, after estimating the variance covariance matrix, you go for using t-test, just like it is being used in usual linear regression models, correct?

josef...@gmail.com

unread,
Dec 7, 2022, 8:23:48 AM12/7/22
to pystat...@googlegroups.com
On Wed, Dec 7, 2022 at 2:12 AM Himadri Manna <himadri...@gmail.com> wrote:
SO, after estimating the variance covariance matrix, you go for using t-test, just like it is being used in usual linear regression models, correct?

Yes, only computation of cov_params is a special method, all other standard inference like p-values, confint, t-test, wald-test are the same as in other models and inherited from a base model.

Josef 
 
Reply all
Reply to author
Forward
0 new messages