Page 1 of 1

Multivariate RP

Posted: Fri May 26, 2006 20:18
by Gerald
I am interested in know about the analysis of multiple variables in time. In other words the recurrences of one variable in relation with others.

My research is related with rainfall and runoff process in hydrology. In principle there are many time relationships between different spatial precipitations and other enviromental variables.

Could someone tell me if there is any research working with analysis combining multiple variables.

Posted: Tue May 30, 2006 07:29
by Norbert
You could utilise joint recurrence plots (JRP), which are real multi-variate tools. Although in publications the JRPs are used as bi-variate tools, do not hesitate to apply them as multi-variate tools.

Have also a look at the comments and suggested literature at http://forum.recurrence-plot.tk/viewtopic.php?t=12

But maybe someone has another idea how to process multi-variate data? I would be curious about it.

Best regards,
Norbert

Re: Multivariate RP

Posted: Fri Oct 28, 2011 17:46
by sultornsanee
I am interested in the multivariate RPs a lot. I have an idea for multiple time series analysis using RPs or RQA. I will make an experiment and give you hints soon. :D

Re: Multivariate RP

Posted: Fri Oct 28, 2011 21:17
by Norbert
Hi Toni,

I look forward to see your approach.

Best regards
Norbert

Re: Multivariate RP

Posted: Tue May 25, 2021 10:18
by Andrii
Dear Norbert,

It was recently highlighted (Wallot, 2019) that the application of multiple JRPs can have the limitation of not catching all the richness of the multiple signals dynamics because the JRPs are highly influenced by the smallest recurrent structures of the individual plots. Apparently, the use of Multidimensional RQA (only in R at the moment) does not have this issue. I was wondering whether there is a modification of JRP that can overcome the limitation of being dominated by the smallest recurrent structure?

Thank you,
Andrii

Re: Multivariate RP

Posted: Wed Aug 4, 2021 21:04
by Norbert
The multivariate RP is nothing else than using the different time series as the different components of the phase space vector. Therefore, there is no difference to the standard approach, as far as I see it.