Whether by differential equations (ODEs or PDEs) or algebraic systems of equations, computational science traffics mainly in deterministic models of physical systems. While we sometimes make the mistake of thinking that the model "is" the physical system, it is always only an abstract, and approximate, representation. Statistical models are a different kind of abstraction, not for the deterministic part of a system, but for how it interacts with physical stochasticity and/or with Bayesian priors expressing imperfect knowledge. This talk compares statistical and deterministic models, using examples drawn from cosmology and bioinformatics.