Measures of the Contribution of Independent Risk Variants to Disease
Alert warning = 1
\(Note:\) Some combinations of K, RAF, RBb and RBBB generate non-estimable results. This usually occurs when RRBB is large, which is usually when RAF is small. If RAF is small, homozygotes of the risk allele are rare and the estimate of RRBB will have wide confidence intervals. Under these circumstances the equation that calculates kbb fails the assumptions of the multiplicative risk model and the probability of disease for risk allele homozygotes, kbb*RRBB, is estimated to be greater than 1. When this occurs we fudge the calculations by reducing RRBB by iteratively reducing RRBB, in up to 10 steps to a minimum of RRBb, until an RRBB is reached that generates a kbb that satisfies kbb*RRBB .< 1. When Alert =1, RRBBcalc outputs the RRBB used in calculations. Since the risk allele homozygotes are rare, the calculations are relatively insensitive to the value of RRBB. We output kbb, kbbRRBb and kbbRRBB (which is kbbRRBBcalc) to aid diagnostics, if these are negative or >1, then there is an input problem and disregard the results.All calculations are underpinned by model assumptions- and acknowledge the old adage: that all models are wrong but some models are useful.
John S. Witte, Peter M. Visscher, Naomi R. Wray