Document worth reading: “Statistical methods research done as science rather than mathematics”

This paper is about how we study statistical methods. As an occasion, it makes use of the random regressions model, whereby the intercept and slope of cluster-specific regression strains are modeled as a bivariate random affect. Maximizing this model’s restricted chance normally gives a boundary worth for the random affect correlation or variances. We argue that this generally is a draw back; that it is a draw back as a results of our self-discipline has little understanding of how updated fashions and methods map info to inferential summaries; that we lack such understanding, even for fashions as simple as this, resulting from a near-exclusive reliance on arithmetic as a way of understanding; and that math alone is no longer ample. We then argue that as a self-discipline, we’re in a position to and can break open our black-box methods by mimicking the 5 steps that molecular biologists usually use to interrupt open Nature’s black bins: design a simple model system, formulate hypotheses using that system, test them in experiments on that system, iterate as wished to reformulate and test hypotheses, and finally test the results in an ‘in vivo’ system. We exhibit this by determining circumstances beneath which the random-regressions restricted probabilities usually tend to be maximized at a boundary worth. Resistance to this technique seems to come back up from a view that it lacks the understanding or psychological heft of arithmetic, perhaps as a results of simulation experiments in our literature infrequently do further than measure a model new method’s working traits in a small fluctuate of situations. We argue that such work may make useful contributions along with, as in molecular biology, the findings themselves and customarily the designs used throughout the 5 steps; that these contributions have as rather a lot wise worth as mathematical outcomes; and that subsequently they benefit publication as rather a lot as the mathematical outcomes our self-discipline esteems so extraordinarily. Statistical methods research done as science rather than arithmetic