In Defense of Fishing

Using an example from animal cognition, I argue that the problems of bias—inherent in choosing null hypotheses or setting Bayesian priors—can sometimes be avoided altogether by collecting more and better observational data before setting up tests of any sort.

Estimation of groundwater recharge using multiple climate models in Bayesian frameworks

Groundwater recharge plays a vital role in replenishing aquifers, sustaining demand, and reducing adverse effects (e.g. land subsidence). In order to manage climate change-induced effects on groundwater dynamics, climate models are increasingly being used to predict current and future recharges. Even though there has been a number of hydrological studies that have averaged climate models’ predictions in a Bayesian framework, few studies have been related to the groundwater recharge. In this study, groundwater recharge estimates from 10 regional climate models (RCMs) are averaged in 12 different Bayesian frameworks with variations of priors. A recession-curve-displacement method was used to compute recharge from measured streamflow data. Two basins of different sizes located in the same water resource region in the USA, the Cedar River Basin and the Rainy River Basin, are selected to illustrate the approach and conduct quantitative analysis. It has been shown that groundwater recharge prediction is affected by the Bayesian priors. The non-Empirical Bayes g-Local-based Bayesian priors result in posterior inclusion probability values that are consistent with the performance of the climate models outside the Bayesian framework. With the proper choice of priors, the Bayesian frameworks can produce good results of groundwater recharge with R2, percent bias error, and Willmott's index of agreement of >0.97, <2%, and >0.97, respectively, in the two basins. The Bayesian framework with an appropriate prior provides opportunity to estimate recharge from multiple climate models.

Cosmic Bayes. Datasets and priors in the hunt for dark energy

Bayesian methods are ubiquitous in contemporary observational cosmology. They enter into three main tasks: (I) cross-checking datasets for consistency; (II) fixing constraints on cosmological parameters; and (III) model selection. This article explores some epistemic limits of using Bayesian methods. The first limit concerns the degree of informativeness of the Bayesian priors and an ensuing methodological tension between task (I) and task (II). The second limit concerns the choice of wide flat priors and related tension between (II) parameter estimation and (III) model selection. The Dark Energy Survey (DES) and its recent Year 1 results illustrate both these limits concerning the use of Bayesianism.