Fits the mixed cumulative incidence function models suggested by Cederkvist et al (2018; https://doi.org/10.1093%2Fbiostatistics%2Fkxx072) which decomposes within cluster dependence of risk and timing.