| dc.description | This thesis addresses the problem of assessing segregation of a quantitative trait under a mixed model of inheritance in non experimental populations, when the trait is linked to some monitored marker. There are two main approaches to the linkage problem in non experimental populations, the difference among them being the kind of information about the model of inheritance which is used in making inferences. Sib-pair methods are based only on identity by descent, i.e., they use partial information about the model, and their inference is conditioned on the identity by descent status of the markers. The alternative approach uses the full model information and, traditionally, the inference for these methods is done through the unconditional analysis of cosegregation, usually requiring Monte Carlo methods to evaluate the likelihood. This thesis presents an extension to one of the techniques (the Morton-MacLean integral) to compute the 'exact' likelihood under mixed models of inheritance when the data is at the bottom of the pedigree, regardless the complexity of such pedigree. With the same arguments used to derive this extension, a Monte Carlo sampling scheme in blocks of sibships is proposed. The idea of making linkage analysis conditional on the marker inheritance vector is also explored, together with some issues related to the computation of the conditional likelihood. | en_US |
