| Abstract: | "The cohort parity analysis (CPA) model of David et al. (1988) is studied formally as a three-state parity-progression table. The general solution is found in a form of convex combination of a finite set of solutions which are described explicitly. A parameterization is suggested for a broad subset of solutions which includes two extreme solutions studied in the original publication and maintains the dimension of the entire set. The CPA solution is also treated as a random variate distributed uniformly on the set of all possible solutions. An algorithm is given for computing the marginal distributions without Monte Carlo simulation." (EXCERPT) |
