Rogatko [1995: Am J Med Genet 59:14-23] has proposed a method for risk prediction with linked markers. The actual implementation of this method required that the analytical forms of likelihood and risk functions be specified. It is impractical to obtain the explicit analytical form of these functions in large phase unknown pedigrees. When large phase unknown pedigrees are encountered, the compound risk can be approximated by a transformation of the discrete distribution obtained by computing the likelihood and risk functions over a grid of points. We propose a method to compute genetic risks when the functional form of the pedigree likelihood is unknown. The method was evaluated using a simple pedigree by comparing the results when functional forms were and were not known. This method was also applied to estimate genetic risks for a single pedigree with nonsyndromal X-linked mental retardation using 3 genetic markers linked to the putative disease gene. Linkage data from an extended pedigree were combined with genome mapping data, and recurrence risk distributions were calculated for members of the pedigree. The results suggest that the proposed method provides accurate risk estimate for genetic diseases. Computer programs are available to apply this method whenever genetic markers are suspected of being linked to a disease gene.