We ran 10 MCMC iterations for ABC model selections and checked convergence by running five instances of each simulation for different random number seeds.
This is done by highlighting in a graph structure the implicit production-consumption relationship between reaction instances of the simulation trajectory.
Each experimental observation is compared to a simulated outcome involving the same measurement protocol so, for example, a heteroplasmy variance estimate of 0.01 derived from a sample of 20 cells is compared to 20 instances of stochastic simulation.
By using the unique identifiers of the species in a reaction trajectory, which implicitly indicate the production-consumption relationship between reaction instances of the simulation, we construct a directed graph structure.
Other studies confirm the power of games and simulations in developing cognition abilities, especially in the instances of virtual simulations enhancing complex cognitive skills (Helle et al., 2011; Siewiorek, 2013), such as self-assessment (Arias Aranda, 2010), or higher-order thinking (Crocco et al., 2016).
In many instances of our simulations, it can be seen as in Figure 3 that the degree distribution is a good estimation for the power-law degree distributions in the real-life social networks.
As discussed above, the performance of different regression methods and training procedures varied widely in our simulated data, with each individual method performing both best and worst in individual instances of 500 simulations from the same population.
We include several task sequences where we first ask students to make a prediction about the outcome of a particular instance of the simulation (after changing one of the initial conditions), then instruct them to run the simulation, observe the outcome, and then to respond to a reflection question.
Upon receipt of a request to execute a simulation, the middleware sets up the proper system environment and then launches an instance of the simulation engine with the appropriate configuration and parameters according to the instructions received from the client.
The intrinsic complexity in the parallelization of the single step has traditionally led to the exploitation of parallelism in the computation of independent instances of the same simulation, which should anyway be computed to achieve statistical convergence of simulated trajectories (as in all Monte Carlo methods).
