Application of Exploratory Data Analysis Methods in Computing Screening Intervalsfor Selected Study Measures of Effectiveness

摘要

This report presents a methodology for the construction of screening intervalsfor selected study measures of effectiveness (MOE): force loss exchange ratio, helicopter, and tank system exchange ratios. Using exploratory data analysis techniques, measures of location and scale are derived from study MOE data. Two methodologies are presented: A test for specious data and a test for comparing two batches of data. The first methodology is based on the biweight estimator of location and the second is based on the median of the batch. The first methodology is to be incorporated into an expert system currently under development at TRADOC Analysis Command-Fort Leavenworth (TRAC-FLVN). The expert systems is designed to fill a quality control need for comparing emerging study results with past-related study MOE data to insure validity of study results. Prediction intervals are proposed for use as a screening tool to determine the acceptability of new MOE data. This will provide statistical validity and quality control to emerging study results. The second methodology is for use when two batches of data are to be compared. Both methodologies will serve as screening tool for the study analyst. (kr)