Thanks in advance for your answers, I'm trying to decompose a stock price using the Rssa package provided in R, after millions of trials, I tried to change the window Length and the eigentriples, the error between the forecast and the test series is really High, as definition the Window length should be chosen between 1 and (N/2) which is the half of time series of length (N), if we multiply (N/2) for periodicity we should obtain the optimal separation of the components, my question now is "How I should decide the Window length if stock prices do not have an explicit periodicity or seasonal components?"

  • $\begingroup$ The window is less important than the choice of kernel. Read what Emmanuel Parzen has to say about the Epinechnikov kernel. $\endgroup$ – Michael Chernick Feb 15 '17 at 14:04
  • $\begingroup$ I've never had to choose the kernel in Rssa, what are you referring to? $\endgroup$ – Daniele Mondi Feb 15 '17 at 14:14
  • $\begingroup$ That is probably because the Epanechnikov kernel is automatically used. $\endgroup$ – Michael Chernick Feb 15 '17 at 14:17
  • $\begingroup$ oh okay i will read the text then, but actually for the window length have you got some tips to use? $\endgroup$ – Daniele Mondi Feb 15 '17 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.