I am trying to fit dshw for the double seasonal times series, and compare the results with tbats() , I used:

> dresid.train.dshw <- dshw(dresid.ts,h = 44)    
Error in dshw(dresid.ts, h = 44) : 
dshw not suitable when data contain zeros or negative numbers

It seems that dshw can not deal with negative or zero values. Does anyone know how to fit dshw on negative data?

Also, from a dshw example, the parameters (alpha, beta, gamma, omega, and phi) are specified in the model. I am wondering how those parameters are selected?

May I also know how to get confidence interval for the dshw forecast? As I can always extract upper and lower from the forecast() of a tbats model, but it seems that I can not find those in the dshw. Thanks.

  • 2
    $\begingroup$ An easy work around would be to add a large constant and run the method, on you obtain the forecast just subtract the constant. $\endgroup$
    – forecaster
    Commented Aug 15, 2016 at 11:21


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.