We compare two algorithms for the numerical estimation of the correlat
ion dimension from a finite set of vectors: the ''classical'' algorith
m of Grassberger and Procaccia (GPA) and the recently proposed algorit
hm of Judd (JA). Data set size requirements and their relations to sys
tematic and statistical errors of the estimates are investigated. It i
s demonstrated that correlation dimensions of the order of 6 can corre
ctly be resolved on the basis of about 100 000 data points in the case
of a continuous trajectory on a strange attractor; the minimum data s
et size is, however, noticeably dependent on the geometrical structure
of the system from which the vectors were sampled. (C) 1998 Elsevier
Science B.V. All rights reserved.