A State-Based Sensitivity Analysis for Distinguishing the Global Importance of Predictor Variables in Artificial Neural Networks
(a) Correctness of state-based sensitivity, local sensitivity, and connected weights analyses (SBSA, local SA, and CWA, resp.) in ranking singular variables within the Case Study, as compared to actual variable importance ranks. (b) Absolute deviations of modeled rankings (arising from SBSA, local SA, and CWA) for variables from actual rankings. Data are means standard error; .