The age at onset distribution was investigated with the maximum likelihood estimation of the finite normal mixture (Kolenikov). We tested multiple mixture models with increasing number of theoretical components and assessed whether this increase improve the fit by using Akaike’s information criterion (AIC) of parsimony. We also computed the Log likelihood value (LLH) for each model, with AIC=−2*log-likelihood+2*k, where k is the number of estimated parameters.We then defined age at onset cut-off points by adding or subtracting one standard deviation from the estimated mean of the theoretical component distribution, following a method described elsewhere (Lin et al., 2006).
