The computational time required to complete a genetic association study of 10 000 SNPs using the Cox model was approximately 18 times greater than the equivalent analysis using logistic regression for cohort studies (cohort size of 10 000), regardless of the cumulative disease incidence (Supplementary Table S7). Similarly, the computational time needed to analyse 10 000 SNPs with the Cox model was at least 190 times greater than the same analysis using logistic regression for case-cohort studies (sampling fraction of 15% initial cohort of 40 000). We expect these relative differences in computational time between the models to remain the same for data sets with greater numbers of SNPs.
