The use of statistical modeling in the estimation of Oringin-Desityation (OD) matrix from trafinc counts is ceviewed In particular modets that consider explicitly the pres-ence of measurement and sampling erors in the observed link fiows are discussed This paper proposes treating the link choice proprtion as random variablcs Accordingly. new statistical models are fomulated and the corresponding Maximum Likelihood Estimator and Bayesiac Estimator of the OD matrix are developed The accuracids of these estimators are compared with those obtainc by prvious mcthods Copyright