NCJ Number
243790
Journal
Polygraph Volume: 42 Issue: 2 Dated: 2013 Pages: 80-111
Date Published
2013
Length
32 pages
Annotation
This study used archival scores to evaluate the multi-facet hypothesis regarding test questions in the Federal ZCT when scored via a rank order test data analysis (TDA) model.
Abstract
Archival scores were used to evaluate the multi-facet hypothesis regarding test questions in the Federal ZCT when scored via a rank order test data analysis (TDA) model. Data were analyzed using bootstrapping and multivariate analysis. Dimensional profiles and statistical confidence intervals were calculated for criterion accuracy using different decision rules. Results from these analyses do not support the multi-facet hypothesis, which worsened criterion accuracy. Criterion accuracy was highest when the rank order scores were interpreted as a single issue. Unweighted combined decision accuracy for grand total scores was 87.6 percent, excluding inconclusive results, and a 12.1-percent inconclusive rate. Results of the rank order model were compared to three other TDA algorithms: OSS-2, OSS-3, and a replication of the Probability Analysis algorithm. All three computer algorithms achieved decision accuracy levels over 90 percent, with inconclusive rates less than 20 percent. Results based on rank order scores were also compared to results from a previous study in a cohort of inexperienced scorers evaluating the same confirmed case sample with the Empirical Scoring System. There was no advantage to the rank order model. Issues surrounding test item variance in rank order and nonparametric models are discussed. Possible improvements to a rank order scoring model were evaluated, and a weighted rank order model achieved an accuracy level of 90 percent, with 19 percent inconclusives. (Published Abstract)