A faster and more general hidden Markov model algorithm for multipoint likelihood calculations

There are two basic algorithms for calculating multipoint linkage likelihoods: in one the computational effort increases linearly with the number of pedigree members and exponentially with the number of markers, in the other the effort increases exponentially with the number of persons but linearly with the number of markers. We describe a faster version of the latter algorithm for which there is no penalty in making the recombination fraction meiosis specific. This can lead to faster and potentially more powerful linkage analysis whenever the number of nonfounder meioses in a pedigree is not too large.