The beat-to-beat heart rate varies considerably over the course of 24 hours. The Computers in Cardiology Challenge of 2002 was based on understanding and modelling the dynamics of the source of this variation. The diverse range of entries for simulating the RR tachogram emphasises the level of ambiguity that exists when attempting to model the cardiovascular system over long time scales. In 2002, a model that combined both deterministic and stochastic descriptions of the data generating process was developed to reproduce both short term variations and long range correlations in RR intervals. In this paper these ideas are extended and new observations of sleep-wake transitions observed during sleep are incorporated. The new model more faithfully represents some of the key empirical characteristics that have been observed in actual recordings of healthy human subjects from the Physionet database. These characteristics include the distribution of RR intervals conditioned on absolute time periods (the time of day) and relative time periods (the current and neighbouring RR intervals). In particular the model emulates the existence of a scale-free power law distribution during wakefulness and the exponential distribution (with a characteristic time scale) of sleep-wake activity during prolonged sleep. Concurrence with real data is demonstrated using detrended fluctuation analysis and multiscale entropy. In order to increase the model's utility for assessing biomedical signal processing methods applied to the RR interval time series, a label for each beat is available as either normal, ectopic or artifactual. Furthermore, a state label (awake, asleep or transitional) is available.