Page 1 of 1

window size

Posted: Tue Sep 30, 2014 08:10
by denise
Hi,
i try to choose the window size for the windowed recurrence quantification analysis of soccer games. I used the method from Dr. Marvan's diploma thesis. I took increasing window size over the data and calculated the RQA values (wmin = 20 and wmax = 800) . For these i computed the average and the standard deviation for each window size. I ploted the values of L against the increasing window size and the standard deviation against the window size also. for each game L reachs a plateau like it is described in the thesis, but there is no plateau (values are still decreasing) for the standard deviations (RR, DET, LAM, ENT). So i know the minimal window size is at the beginning of the plateau of L, but what i didn't know is how big i can choose the window? Would you call that plateaus? Does anyone have a hint for me? Would be nice. :)

this is one of my plots:
Window size.png
Window size.png (10.89 KiB) Viewed 18681 times
Best regards
Denise

Re: window size

Posted: Tue Sep 30, 2014 14:48
by Norbert
Hello Denise,

the results depend on the window size and also on the length of the entire time series, i.e., how many windows you finally will have. This would need some consideration from your side.

Regarding the plot and the plateaus: I would not really call them plateaus, but after window size 150 the fast decrease of the curves changes to a very slow decrease. Therefore, this value can be considered equivalent as the beginning of the plateau in L.

Hope this helps!

Best
Norbert

Re: window size

Posted: Wed Oct 1, 2014 12:39
by denise
Hello Norbert,

thanks for your reply. It helps :)

Regards!
Denise

Re: window size

Posted: Tue Oct 7, 2014 11:38
by holistic
Hi @ll,

can you please tell me, where to find this method in the Thesis? Also..is it still applied?

Or are there any new methods for determining an "optimal" window size or some general recommendations respectively?

Thanks :)

Re: window size

Posted: Tue Oct 7, 2014 12:01
by Norbert
it is in my quite old diploma thesis in the appendix B (pp. 110) – but only in German.
available at http://recurrence-plot.tk/furtherreading.php

i have not further considered this issue but would clearly be worth to do it.

Re: window size

Posted: Tue Oct 7, 2014 13:43
by holistic
Got it thanks! Deutsch ist kein Problem ;)