Non-uniform sampling of time series data

General discussions and questions about recurrence plot and recurrence network related methods.
Post Reply
dangus
Junior
Posts: 1
Joined: Wed Aug 10, 2011 09:57
Affiliation (Univ., Inst., Dept.): University of Queensland
Location: Brisbane, Australia
Research field: Artificial Intelligence

Non-uniform sampling of time series data

Post by dangus »

Hi all,

I was wondering if anybody knows of software to create a recurrence plot of data that is coded with a time point (in milliseconds), and a value from that time point. The software I have found so far assumes that the time series is uniformly sampled as far as I can tell. An example of my data is:

Time, Amplitude
11234, 9.6
11398, 10.2
12001, 1.7
12121, 2.0

I'd like to create a recurrence plot indicating the difference between the amplitude values, and also one where if a later occurring sample has a greater value than an earlier occurring sample that it is indicated as 1, and 0 otherwise.

Thanks,

Dan
User avatar
sultornsanee
Senior
Posts: 29
Joined: Tue Oct 12, 2010 21:05
Affiliation (Univ., Inst., Dept.): Northeastern University
Location: Boston, Massachusetts, USA
Research field: Pattern Recognition Analysis
Location: Network/Nano Science and Engineering Laboratory, Northeastern University, Boston, USA

Re: Non-uniform sampling of time series data

Post by sultornsanee »

You can download CRP toolbox (Dr. Norbert Marwan)
Sivarit Sultornsanee (Tony)

Network/Nano Science and Engineering Laboratory (NSE Lab)
360 Huntington Ave., Room 320 SN, Boston, MA 02038, USA
E-mail: SULTORNSANEE.S@HUSKY.NEU.EDU
User avatar
Norbert
Expert
Posts: 194
Joined: Wed Jan 4, 2006 11:03
Affiliation (Univ., Inst., Dept.): Potsdam Institute for Climate Impact Research, Germany
Location: Potsdam, Germany
Location: Potsdam Institute for Climate Impact Research, Germany

Re: Non-uniform sampling of time series data

Post by Norbert »

Although the CRP toolbox allows to provide a certain vector with arbitrary time scale, finally, the plot of the matrix is equally sampled. Unfortunately.

But the solution is to use sparse matrices in Matlab and plot the RP using the spy command. This would allow non-equally sampled RPs. See here for an example how to use sparse/spy:
http://www.agnld.uni-potsdam.de/~marwan ... #long_data

However, I'm a bit concerned about this procedure. If we have larger distances, should then the recurrence point in the RP not be larger (covering a larger area in the plot)?
Post Reply