dear all, the points made by Allin and Sven are correct: if feasible, one takes all subsamples of a given length b out of a time series of length n, so there is nothing random. If the series is really long (super high frequency data) one may need to take only some of the possible subsamples,  and that has to be a random choice bye Stefano -- _________________________________________________________________________ Stefano Fachin Professore Ordinario di Statistica Economica Dip. di Scienze Statistiche Università di Roma "La Sapienza" P.le A. Moro 5 - 00185 Roma - Italia Tel. +39-06-49910834 fax +39-06-49910072 web http://stefanofachin.site.uniroma1.it/