The computational time needed to perform Cox and logistic regression models was estimated using 10 000 variants (with RAFs between 0.05 and 0.95) assuming only one of which is causal with an HR between 1 and 2. In the cohort setting, we used 10 000 simulated individuals. For the case-cohort setting, we generated the case-cohort in the same way described above using a sampling fraction of 15% and an initial cohort of 40 000 simulated individuals. We used the survey follow-up model (see above) and compared the computational times of the models for cumulative disease incidences of 5, 10 and 15%. To estimate these computational times we extended GenABEL13 to enable it to perform Prentice-weighted Cox models. This extension is available on request from the authors.
