Dear Karolina,

The despiking option applies a continuous "squashing" function to the BOLD signal (hyperbolic tangent) rather than relying on an absolute threshold and cropping any values above that threshold, so there is no concept of specific points which are despiked vs. points which are not (roughly speaking, for points that lie within +-3 standard deviations around the mean value the squashing function is approximately linear, while for points that lie beyond that the amount of compression becomes increasingly harsher). To clarify the associated formula is:

 a = median(X);
 b = 4*median(abs(X-a));
 Xnew = a + b*tanh((X-a)/b); 

In general, I typically find scrubbing considerably better suited to remove the influence of potential outlier scans compared to despiking. When using scrubbing you can find out if you want the number of data points that have been scrubbed for each subject, for example, by going to the Setup.CovariatesFirstlevel tab, selecting the 'scrubbing' covariate, and clikcing on the 'covariate tools. subject-level aggregate' option. Select there the 'sum' measure in order to compute the total number of data points / scans that have been identified for each subject as potential outlier scans by ART.

Hope this helps
Alfonso
Originally posted by Karolina Finc: