I have been trying to use the rugarch package but I find it sometimes limiting. After certain amount of data points the package doesn't converge and it becomes kind of annoying. Is there any R/Python/Matlab alternative to it? I am trying to fit an apARCH model but it seems like there are not too many options. I also thought about using fGarch, but it also has some disadvantages.

  • 2
    $\begingroup$ With respect to the rugarch package, have you tried setting the solver method to 'hybrid', eg. solver = hybrid? Other than that, you can work with the solver control to try and establish convergence, eg. solver.control = list(tol = 1e-12), gives you a smaller tolerance for the solver? Sometimes the solver fails to converge due to bad initial values? Maybe some of these tips will help you out? :-) $\endgroup$ – Pleb Mar 25 at 15:58
  • $\begingroup$ I tried using the method hybrid. I'll check the tolerance of the solver. I'll give it a try. I also tried optimizing the parameters with the fGarch package and using them as "fixed" parameters on rugarch, but it was hopeless. $\endgroup$ – Francisco Javier Jara Ávila Mar 25 at 19:58
  • 1
    $\begingroup$ I found an alternative on Python, but it still has the nuisance that modelling the mean is not very straightforward. It is particularly a problem if you want to use ARFIMA modelling for the mean. My take on this: package is a good option but right now it is quite limited. More information can be found here $\endgroup$ – Francisco Javier Jara Ávila Mar 26 at 9:39
  • 1
    $\begingroup$ Save yourself the headache and write the code yourself. If you walk into problems, ask the question on stackoverflow ;). $\endgroup$ – simsalabim Mar 29 at 14:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.