A transformation of Bayesian statistics:Computation, prediction, and rationality

Bayesian approaches have long been a small minority group in scientific practice, but quickly acquired a high level of popularity since the 1990s. This paper shall describe and analyze this turn. I argue that the success of Bayesian approaches hinges on computational methods that make a class of models predictive that would otherwise lack practical relevance. Philosophically, however, this orientation toward prediction comes at a price. The new computational approaches change Bayesian rationality in an important way. Namely, they undercut the interpretation of priors, turning them from an expression of beliefs held prior to new evidence into an adjustable parameter that can be manipulated flexibly by computational machinery. Thus, in the case of Bayes, one can see a coevolution of computing technology, an exploratory-iterative mode of prediction, and the conception of rationality.

Uncovering drivers of dose-dependence and individual variation in malaria infection outcomes

To understand why some hosts get sicker than others from the same type of infection, it is essential to explain how key processes, such as host responses to infection and parasite growth, are influenced by various biotic and abiotic factors. In many disease systems, the initial infection dose impacts host morbidity and mortality. To explore drivers of dose-dependence and individual variation in infection outcomes, we devised a mathematical model of malaria infection that allowed host and parasite traits to be linear functions (reaction norms) of the initial dose. We fitted the model, using a hierarchical Bayesian approach, to experimental time-series data of acute Plasmodium chabaudi infection across doses spanning seven orders of magnitude. We found evidence for both dose-dependent facilitation and debilitation of host responses. Most importantly, increasing dose reduced the strength of activation of indiscriminate host clearance of red blood cells while increasing the half-life of that response, leading to the maximal response at an intermediate dose. We also explored the causes of diverse infection outcomes across replicate mice receiving the same dose. Besides random noise in the injected dose, we found variation in peak parasite load was due to unobserved individual variation in host responses to clear infected cells. Individual variation in anaemia was likely driven by random variation in parasite burst size, which is linked to the rate of host cells lost to malaria infection. General host vigour in the absence of infection was also correlated with host health during malaria infection. Our work demonstrates that the reaction norm approach provides a useful quantitative framework for examining the impact of a continuous external factor on within-host infection processes.