This paper deals with identifying finite dimensional normalized right coprime factors (NRCFs) under a stochastic framework from a sequence of their frequency response estimates. A two-step procedure is suggested for the estimation in which two cost functions are sequentially minimized. These cost functions are constructed on the basis of weighted curve fitting and the conditions on NRCFs. A numerical procedure is proposed for computing the estimate. Moreover, it is proved that under some conditions, the estimate converges with probability 1 (w.p.1) to its actual value with increasing the data length, and in general the optimal estimate is asymptotically approximately Gaussian. The efficiency of the proposed method in NRCF identification is illustrated by numerical simulations. (C) 2004 Elsevier Ltd. All rights reserved.