The FAMILY Procedure

Creating Allelic Transmission Scores

Abecasis, Cookson, and Cardon (2000) define a pair of orthogonal allelic transmission scores, $b$ and $w$, the expected genotype and deviate, respectively, for each individual at each marker. To create these scores, the genotype in terms of allele $M_ v$ must first be defined as $g_{jv} = m_{jv}-1$ for individual $j$, where $m_{jv}$ represents the number of $M_ v$ alleles that the genotype comprises. For any founder $j$, an individual whose parents are not observed, in the sample, these scores are defined as $b_{jv}=g_{jv}$ and $w_{jv}=0$. Otherwise, let $M_ j$ and $F_ j$ be the respective indices of the mother and father of individual $j$. Then for any nonfounder, assuming scores for an individual’s ancestors are calculated before his or her own,

\[  b_{jv} = \left\{  \begin{array}{ll} (b_{M_ j v} + b_{F_ j v})/2 , &  b_{M_ j v} \mbox{ and } b_{F_ j v} \mbox{ are nonmissing} \\ \sum _{k\in S_ j} g_{kv}/|S_ j|, &  \mbox{otherwise} \end{array} \right.  \]

where $S_ j=\{ k:M_ k=M_ j, F_ k=F_ j, \mbox{ and }k \mbox{ genotyped}\} $, and then $w_{jv} = g_{jv} - b_{jv}$. These scores are calculated for all alleles at the markers specified in the VAR statement and are included in the OUTQ= data set.