quantile regression using copulas

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

quantile regression using copulas

indus
Hi all,

Has anyone used the qua.regressCOP2 function from the copBasic package???
The default copula function used in this function is plackett copula and I
wanted to use archimedean copula. Attached below is my code:

mycop<-frankCopula
V=seq(0.001,0.99,by=0.000217)
R<-qua.regressCOP2(0.25,V,cop=mycop,para=c(3.504))

And this is the error I get:
Warning messages:
1: In qua.regressCOP2(0.25, V, cop = mycop, para = c(3.504)) :
  could not uniroot in derCOPinv2, skipping sample for i=4558 having
V=0.989869
2: In qua.regressCOP2(0.25, V, cop = mycop, para = c(3.504)) : 3.504


Would really appreciate if anyone could tell me what is causing the problem
and
how I should solve it.

indu

        [[alternative HTML version deleted]]

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Reply | Threaded
Open this post in threaded view
|

Re: quantile regression using copulas

Marius Hofert-3
Please note:

1) your example is not working in the way you provided it (see
http://www.minimalbeispiel.de/mini-en.html)
2) you receive a warning, not an error
3) I'd try and debug qua.regressCOP2 to see why the warning appears
4) in case 3) does not help, contact the maintainer of copBasic (William
H. Asquith <william.asquith at ttu.edu>)

Having said the above, the warning message "... could not uniroot..." is indeed
weak. I guess (without looking at the code) that derCOPinv2 computes some kind
of inverse of the derivative of a copula with respect to one argument. This
often appears since it is the quantile function of a conditional copula. If the
inversion is done via uniroot, it might not be numerically 100% stable and you
probably ran into a case where something strange happened there (not too strange
though, since it's just a warning). But if you debug it (use
debug(qua.regressCOP2) and call your example again) and find the flaw, you can
give the maintainer feedback how the function can be improved so that you won't
run into this problem again in the future.

Cheers,

Marius

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
Reply | Threaded
Open this post in threaded view
|

Re: quantile regression using copulas

indus
Hi Marius,

I have tried debugging the qua.regressCOP2 function.
The error I'am getting is:
"Error in cop(u, v + delv, ...) : unused argument(s) (v + delv)".
Unable to decipher it.
And have mailed to  william.asquith at ttu.edu>.

Thanks
indu