TRENTOOL ERROR: Not enough points in timeseries for current analysis settings

If TEprepare throws the following error

TRENTOOL ERROR: Not enough points in timeseries for current analysis settings: cfg.repPred too big, or max(cfg.ragdim)*max(cfg.ragtaurange) too big, or maximum u to big

the input time series or specified time of interest (toi) is too short for the parameters you've set in cfgTEP, namely the time series is too short to accommodate the requested values for embedding parameter optimization with the Ragwitz criterion.

The relevant parameters are actthrvalue, ragdim, ragtaurange, and repPred - the Ragwitz criterion will test all possible combinations of values in ragdim and ragtau as a potential delay embedding. Here a problem occurs if the longest embedding requested by the given values is too long to fit into the time series. The longest embedding in samples is calculated as:

dim = max(cfgTEP.ragdim);    % specified in cfgTEP.ragdim
tau_factor = max(cfgTEP.ragtaurange);  % specified in cfgTEP.ragtaurange
act = min([cfgTEP.actthrvalue act_data]);  % threshold the ACT calculated from the data
tau = act*tau_factor

If in this example the actual values for dim were 11 and for tau 20, the embedding had a length of (dim-1)*tau=200 samples. If the value for repPred were set to 500, TEprepare would try to call Ragwitz with an embedding of length 200 for the 500 first embedded data points. This would require 700 or more data points in the time series to be analyzed. If the time series is shorter than that, TEprepare throws the error shown above.

Possible solutions are:

- increasing the toi/length of time series to be analyzed
- reduction of any of the four parameters `cfgTEP.ragtaurange`, `cfgTEP.ragdim`, `cfgTEP.repPred`, `cfgTEP.actthrvalue`
- checking the actual ACTs in the data by calling 'act = TEgetACT(cfgTEP,data);' - the maximum ACT found over trials may be due to an outlier, the `actthrvalue` can be set to the mean ACT plus 2 SD or some other criterion for outlier detection

---

Go back to the FAQ page.