Premium
Measures of association for comparing analytical methods generating ordinal results
Author(s) -
Gerlach Robert W.
Publication year - 1998
Publication title -
journal of chemometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.47
H-Index - 92
eISSN - 1099-128X
pISSN - 0886-9383
DOI - 10.1002/(sici)1099-128x(199803/04)12:2<105::aid-cem503>3.0.co;2-i
Subject(s) - categorical variable , rank correlation , concordance , ordinal data , statistics , spearman's rank correlation coefficient , association (psychology) , correlation coefficient , correlation , concordance correlation coefficient , rank (graph theory) , mathematics , correlation ratio , ordinal scale , econometrics , data mining , computer science , psychology , medicine , geometry , combinatorics , psychotherapist
Abstract The use of various measures of association that rate the agreement of one chemical analysis method with another has generally been limited to statistics comparing results based on continuous data types. This paper introduces a number of measures of association more commonly applied in medical or biological research and evaluates their ability to serve as figures of merit for summarizing the performance of semiquantitative methods. The measures of association are applied to actual and synthetic data sets to demonstrate responsiveness to changes in level of agreement. Only measures of association that preserved rank information provided the sensitivity required in a method comparison. Somers’ d , Kendall's τ b and rank correlation coefficients were found to perform well in the tested scenarios. The concordance correlation coefficient and Spearman's rank correlation coefficient were identified as the best overall measures of association for summarizing ordinal categorical method performance. The concordance correlation coefficient is recommended if significant bias is present. © 1998 John Wiley & Sons, Ltd.