The capacity to generate lots of data has become the hallmark of current biomedical science, and with this excess of results investigators have turned to bioinformatic specialists or statisticians to make sense of them. Artificial intelligence (including machine learning) and advanced statistics (especially predictive modeling and causal inference) are used to convert the surfeit of data into discernible trends, but they also act to remove the investigators from the original data since the algorithms and computations of these analytical tools are opaque. The idea is presented that analytical procedures which are impossible to assess take on the characteristics of a faith-based paradigm which is incompatible with the scientific method. In this chapter I also argue against the idea that more data are preferable to less data and suggest a focus on iterative analysis, an alternative approach informed by Bayesian statistics involving conditional probability.