Statistical Measurement of Trees' Similarity
Diagnostic theories are fundamental to Information Systems practice and are represented in trees. One way of creating diagnostic trees is by employing independent experts to construct such trees and compare them. However, good measures of similarity to compare diagnostic trees have not been identified. This paper presents an analysis of the suitability of various measures of association to determine the similarity of two diagnostic trees using bootstrap simulations. We find that three measures of association, Goodman and Kruskal's Lambda, Cohen's Kappa, and Goodman and Kruskal's Gamma (J Am Stat Assoc 49(268):732—764, 1954) each behave differently depending on what is inconsistent between the two trees thus providing both measures for assessing alignment between two trees developed by independent experts as well as identifying the causes of the differences.
Sabbaghan, S., Chua, C. E., & Gardner, L. A. (2020). Statistical Measurement of Trees' Similarity. Quality and Quantity Springer.
The definitive version is available at https://doi.org/10.1007/s11135-019-00957-8
Business and Information Technology
Keywords and Phrases
Diagnostic theory; Threshold building; Tree
International Standard Serial Number (ISSN)
Article - Journal
© 2020 The Authors, All rights reserved.
01 Jan 2020