Hazard and probabilities of unknown genotypes

The likelihood of a genotype of an unsampled individual depends upon genetic data from relatives. When these relatives may be carriers of a mutation as in X-linked disease, the likelihood of a genotype is altered by the consequences of the possible genotypes on offspring. The probabilities of the possible genotypes are easily derived from the probabilities conditional on the genotype obtained from a computer program aimed at risk calculation. This procedure of genotype reconstruction brings to light a potential hazard in the interpretation of computer data: a compound risk may be better represented as the harmonic mean of a set of conditional probabilities than the simple average of these probabilities.