Page 1 of 1

Short data segments

PostPosted: Thu Mar 23, 2006 14:55
by geo
Dear all,

In some articles on RP's I often read that the method is also suitable for short data segments of non stationary type (cfr Marwan, Saparin and Kurths "Generalisation of Recurrence Plot Analysis for Spatial Data" 2005 Intl Symp on Nonlinear Theory and Its applications Bruges belgium oct 18=25, 2005. ).
This is of course a very attractive statement but what is "short" ? Surely, the data segment should be long enough for the attractor to become deployed in the phase space ? I will try some test on an EEG data segment ofr different length. Or has this already been done ?

Sincerely,

DR. G. Otte

PostPosted: Tue Mar 28, 2006 10:43
by Norbert
it depends what or how much you are interested on. if you are interested to derive dynamical invariants you will need longer data series. however, if you are working with measures of complexity, as the RQA measures are, and your time series covers e.g. transitions, data lengths of 200 - 1000 could be enough.

best regards,
norbert