Hyunsook Hong1, Yunhee Choi2, Seokyung Hahn3, Sue Kyung Park4, Byung-Joo Park5. 1. Division of Medical Statistics, Medical Research Collaborating Center, Seoul National University College of Medicine, Seoul National University Hospital, Seoul, Korea; Department of Biostatistics, Seoul National University School of Public Health, Seoul, Korea. 2. Division of Medical Statistics, Medical Research Collaborating Center, Seoul National University College of Medicine, Seoul National University Hospital, Seoul, Korea. Electronic address: yhchoi@snuh.org. 3. Division of Medical Statistics, Medical Research Collaborating Center, Seoul National University College of Medicine, Seoul National University Hospital, Seoul, Korea; Department of Medicine, Seoul National University College of Medicine, Seoul, Korea. 4. Department of Preventive Medicine, Seoul National University College of Medicine, Seoul, Korea; Department of Biomedical Science, Seoul National University Graduate School, Seoul, Korea; Cancer Research Institute, Seoul National University College of Medicine, Seoul, Korea. 5. Department of Preventive Medicine, Seoul National University College of Medicine, Seoul, Korea.
Abstract
PURPOSE: Kappa is a widely used measure of agreement. However, it may not be straightforward in some situation such as sample size calculation due to the kappa paradox: high agreement but low kappa. Hence, it seems reasonable in sample size calculation that the level of agreement under a certain marginal prevalence is considered in terms of a simple proportion of agreement rather than a kappa value. Therefore, sample size formulae and nomograms using a simple proportion of agreement rather than a kappa under certain marginal prevalences are proposed. METHODS: A sample size formula was derived using the kappa statistic under the common correlation model and goodness-of-fit statistic. The nomogram for the sample size formula was developed using SAS 9.3. RESULTS: The sample size formulae using a simple proportion of agreement instead of a kappa statistic and nomograms to eliminate the inconvenience of using a mathematical formula were produced. CONCLUSIONS: A nomogram for sample size calculation with a simple proportion of agreement should be useful in the planning stages when the focus of interest is on testing the hypothesis of interobserver agreement involving two raters and nominal outcome measures.
PURPOSE: Kappa is a widely used measure of agreement. However, it may not be straightforward in some situation such as sample size calculation due to the kappa paradox: high agreement but low kappa. Hence, it seems reasonable in sample size calculation that the level of agreement under a certain marginal prevalence is considered in terms of a simple proportion of agreement rather than a kappa value. Therefore, sample size formulae and nomograms using a simple proportion of agreement rather than a kappa under certain marginal prevalences are proposed. METHODS: A sample size formula was derived using the kappa statistic under the common correlation model and goodness-of-fit statistic. The nomogram for the sample size formula was developed using SAS 9.3. RESULTS: The sample size formulae using a simple proportion of agreement instead of a kappa statistic and nomograms to eliminate the inconvenience of using a mathematical formula were produced. CONCLUSIONS: A nomogram for sample size calculation with a simple proportion of agreement should be useful in the planning stages when the focus of interest is on testing the hypothesis of interobserver agreement involving two raters and nominal outcome measures.
Authors: Yoseba Cánovas Zaldúa; Bartomeu Casabella Abril; Carlos Martín Cantera; Fernando González García; Sonia Moreno Escribá; José Luis Del Val García Journal: Aten Primaria Date: 2019-01-10 Impact factor: 1.137