Depiction of Certainty of Exploratory versus Confirmatory Approaches. A depiction of the research process with greater uncertainty to the left and greater certainty to the right. Fully unsupervised machine learning (where no relationship between variables is clear) should be placed entirely on the left. Supervised machine learning, in which the “dependent variable” (or signal) is clear and certain predictor variables are likely well specified (such as in our case) fall somewhat to the right of that. Split-half cross-validation (where datasets are split in two and explored and confirmed) fall yet further to the right. The “most confirmatory approach” is a “close” or direct replication.