The affected sib methods, which are used to make inferences about the genetic components of HLA associated diseases, have many underlying assumptions which may not always be realistic. These include no selective disadvantage of affected individuals, little or no recombination between the marker loci and the 'disease' locus, a single panmictic population, Mendelian segregation of the disease locus alleles and random distribution of individuals over environments. The effects of breaking these assumptions have been investigated. We have explicitly derived the haplotype sharing identity by descent (IBD) expectations for the cases of selection against affected individuals and recombination between the HLA marker loci and the 'disease' predisposing locus for affected sib trios (as was previously done for affected sib pairs). We have also derived, for both affected sib pairs and trios, the haplotype sharing expectations for non-random mating (positive assortative), admixture, meiotic drive (of disease allele carrying haplotypes), and a random versus shared environmental component for sibs. In order to assess the sensitivity of the affected sib methods to perturbations in the assumptions, the expectation spaces of haplotype sharing in affected sib pairs and sib trios under the single diallelic locus model with varying penetrances and allele frequencies are fully described. The effects on haplotype sharing and subsequent disease parameter estimation are different for each of the factors we have considered. The affected sib methods are found to be robust in many situations.