Empirical power analysis showed that our sample affords 45.3 and 87.4 % power to detect GVs explaining 1 and 1.5 % phenotypic variance, respectively (genomewide alpha = 1 × 10−8). Relative to the logistic model, the survival model is expected to show superior power especially for locating low frequency causal GVs (see e.g., van der Net et al. 2008). However, the above power computations are informative also for the age at onset phenotype given the large overlap among the samples included in the two analyses and the slightly lower size of the sample we used in the survival analysis.
