Model-free (nonparametric) linkage relies only on allele sharing by first-degree relatives, as opposed to parametric linkage which requires precise specification of the mode of trait inheritance. In this approach, affected sibling pairs are usually studied [24]. If genetic variation at a locus plays a role in determination of a phenotype, then the effected siblings will share alleles or genetic variations at the locus more often than this is predicted by chance [63]. These methods can cope with variable penetrance and do not require parental genotype, which is an advantage in disorders of late onset like hypertension [24]. However, this approach may rise statistical power issues as many areas of the genome may be searched for cosegregation of alleles; therefore, a correction must be made for multiple comparisons [24, 64]. Accordingly, it has been argued that a stringent significance level for a definite nonparametric linkage should be set at p < 0.00001 Õ 2.2, to reduce the risk of false-positives [63]. Large collections of affected sibling pairs would therefore be needed to offer adequate power to fulfil these criteria [64].
© 2001 Alexander Binder