UniODA vs. McNemar’s Test for Correlated Proportions: Diagnosis of Disease Before vs. After Treatment

Paul R. Yarnold

Optimal Data Analysis, LLC

McNemar’s test is used to assess the significance of the difference between two correlated proportions: for example when the two proportions are based on the same sample of subjects or on matched-pair samples. This methodology is compared with UniODA using two examples in disease diagnosis before vs. after treatment.

View journal article

Estimating Inter-Rater Reliability Using Pooled Data Induces Paradoxical Confounding: An Example Involving Emergency Severity Index Triage Ratings

Paul R. Yarnold

Optimal Data Analysis, LLC

The inter-rater reliability of patient triage ratings made using the Emergency Severity Index is computed and compared for pooled data within and across studies versus separately for pairs of independent raters. Findings reveal that results for the pooled findings exhibit confounding attributable to Simpson’s Paradox. These findings raise the concern that all estimates of inter-rater (and parallel-forms) reliability based on pooled data reported for all instruments discussed in the literature are susceptible to paradoxical confounding, and likely result in overestimation of rating reliability.

View journal article