This talk presents a developing programme in meta-research, or research about research, in which we simulate data to test hypotheses about medical research and clinical practice. Simulated data has a number of advantages over real data: access to real data is often impossible, especially in clinical science; we do not have unfiltered access to the truth of hypotheses with real data, while simulations allow us to stipulate the truth; and real data does not permit counterfactual analyses, while simulations afford the exploration of many counterfactual scenarios. In one set of simulations we investigate features of trials, asking questions like: to what extent does publication bias exaggerate estimates of effectiveness? In another, particularly important for precision health, we model the patient-physician encounter and ask: how reliable are first-person inferences about drug effectiveness?