An Employing Receiver Operating Characteristic (ROC), Probability Sensitivity and Specificity to Determine Significant Influencers

Abstract

Estimates may be susceptible to uncertainty in the sensitivity and specificity of the test when determining the existence of a challenge. Bayesian inference is a logical method of propagating these uncertainties, whereas hierarchical modelling captures the variation of these parameters across experiments. These two techniques are employed in conjunction with Bayesian inference. Statistic adjustments cannot compensate for selection bias in a sample without making significant assumptions; however, multilevel regression and poststratification can at least account for known differences between models. Finding essential drivers in regression modelling through the use of Bayesian sensitivity-specificity and the receiver-employing feature is recommended in this paper, and findings that are easily interpretable are obtained as a result. The results of numerical evaluations of various methodologies indicate that this approach may help implement the principles of statistical modelling and assessment, thereby aiding scientists and executives in the process of making meaningful decisions.