Chi-square tests, ANOVA, and tetrachoric correlations were conducted to examine variation in nicotine withdrawal symptoms and related-smoking characteristics across the Australian and Finnish samples using Stata (2003), which adjusts standard errors for the non-independence of measures within family members. Multipoint and single-point non-parametric linkage was conducted in MERLIN (Abecasis et al., 2002), first separately in each sample and then after combining both samples. In the combined analyses, allele assignments and frequencies were calculated separately for the Australian and Finnish samples using a slight modification to the program’s source code (G. Abecasis, personal communication; see Saccone et al., 2007a). For the multi-point linkage analyses, the 2cM grid option was used. Evidence for linkage was supported in a particular genomic region, if the number of alleles shared identical by descent was greater than would be expected by chance in sib-pairs concordant for DSM-IV nicotine withdrawal. For LOD scores greater than 3.0, we obtained genome-wide adjusted p-values using MERLIN both to generate 1000 simulated datasets and to test that data for linkage. This provided a threshold for determining the degree to which we would obtain such a LOD score due to just chance alone.
