Many dynamical systems can successfully be analyzed using the temporal network formalism. This is notably the case for the human interaction networks that support the spread of contagion and information processes in the population. Empirical temporal networks and dynamic processes that take place in these situations show heterogeneous, non-Markovian, and intrinsically correlated dynamics, making their analysis particularly challenging. Randomized reference models (RRMs) constitute a versatile toolbox for studying such systems. Defined as ensembles of random networks with given features constrained to match those of an input (empirical) network, they may be used to identify statistically significant features in empirical temporal networks (i.e.\ different from the null random networks) and to infer the effects of such features on dynamical processes unfolding in the network. However, the effects of most randomization procedures on temporal network features remain poorly understood, rendering their use non-trivial and susceptible to misinterpretation.
Here we propose a unified framework for classifying and understanding microcanonical RRMs (MRRMs), which constrain chosen features to take exactly the same value as in the empirical network but are otherwise random. The framework lets us order MRRMs and deduce their effects on important temporal network features, and we use it to show how we may generate new MRRMs from existing ones by sequential composition of independent MRRMs. We show how to apply the framework to unravel how different features of an empirical network of mobile-phone calls influence the spread of information.