PubMed Health⌕ Search
Home/Archive/PubMed

PubMed · 1424582

Evaluating agreement between two methods for measuring the same quantity: a response.

J Lee
View original source ↗Export metadata ↓

Abstract

A method comparison study is typically concerned with evaluating the agreement between a "new" method against an "established" method for measuring the same quantity. Various statistical indices have been advocated for measuring the extent to which two methods give identical readings, that is, absolute method agreement. This article scrutinizes two contending statistical indices, the intraclass correlation and the "limits of agreement", for measuring absolute method agreement, and found neither index to be without deficiency. Notwithstanding, each index can serve a useful role in quantitating method agreement in carefully considered situations.

Explore related subjects

Reproducibility of ResultsStatistics as TopicComparative StudyJournal Article

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

J Lee. 1992. Evaluating agreement between two methods for measuring the same quantity: a response.. https://doi.org/10.1016/0010-4825(92)90025-i

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed2026

An Assessment of Reliability Estimation Methods for Binomial Health Care Quality Measures.

We evaluated the performance of commonly used methods for estimating the reliability of binomial health care quality measures using simulated datasets spanning a range of performance score means and variances, numbers of entities, and patient sample sizes. For each simulation, reliability was estimated for all selected methods and compared with the known true reliability derived from the simulation parameters, with methods assessed on their accuracy and precision. Logistic regression with reliability estimated on the outcome scale demonstrated the highest accuracy and precision among all methods evaluated. The widely used Adams beta-binomial method performed poorly, although a modification recommended by Nieser and Harris substantially improved its performance. These approaches are applicable only to binomial measures. Among methods that can be applied to both binomial and continuous measures, permutation resampling of the Spearman rank correlation coefficient was the most accurate and precise, outperforming other commonly used approaches. Overall, for binomial quality measures, logistic regression on the outcome scale is the preferred method for reliability estimation, followed closely by the modified beta-binomial approach, while for non-binomial measures, permutation-based Spearman rank correlation appears to be the most suitable method.

Reproducibility of Results↗