Statistical mechanics is fundamentally based on calculating the probabilities of molecular-scale events. Although Bayes’ theorem has generally been recognized as providing key guiding principles for setup and analysis of statistical experiments, classical frequentist models still predominate in the world of computational sciences. In this presentation, I’ll talk about how we got around some interpretation issues in the way of making predictions from our simulation data. First, we present a method of calculating the error in data derived from counting simulation outcomes. Application of this method to counting molecular cavity sizes in water has led to an increased understanding of solvation free energies. Next, we show how whole functions can be estimated in a robust way based on observations of independent and dependent variables. This result has been used to accurately parametrize Langevin dynamics simulations of coarse-grained molecular models.
