TY - JOUR

T1 - The Impact of Sample Size on Reliability Metrics Stability in Isokinetic Strength Assessments

T2 - Does Size Matter?

AU - Warneke, Konstantin

AU - Keiner, Michael

AU - Wallot, Sebastian

AU - Siegel, Stanislav D.

AU - Günther, Christian

AU - Wirth, Klaus

AU - Puschkasch-Möck, Sebastian

N1 - Publisher Copyright: © 2025 The Author(s). Published with license by Taylor & Francis Group, LLC.

PY - 2025/4/23

Y1 - 2025/4/23

N2 - The ability to reliably capture performance parameters must be considered as crucially important to produce valid study results. The ICC and the inclusion of the calculation of the standard error of measurement and the minimal detectable change became the most common way to justify subsequent testing procedures to be reliable. However, early studies around the new millennium identified weaknesses of the ICC and proposed the implementation of more elaborate procedures, including the quantification of the systematic bias and the quantification of the random error via the mean absolute error or mean absolute percentage error. According to the law of large number and earlier research indicating that relative indices such as correlation coefficients necessitate a minimum sample size to stabilize, it was hypothesized that reliability indices follow an optimal sample size trend. In accordance with previous studies in correlation coefficients, this study highlights the importance of including high numbers of participants to receive stable reliability measures. The random error was not significantly affected by increased samples while providing important information about the performed standardization success in the testing, the study also underlines the relevance of reporting not only ICC-based reliability statistics but also the quantification of random errors.

AB - The ability to reliably capture performance parameters must be considered as crucially important to produce valid study results. The ICC and the inclusion of the calculation of the standard error of measurement and the minimal detectable change became the most common way to justify subsequent testing procedures to be reliable. However, early studies around the new millennium identified weaknesses of the ICC and proposed the implementation of more elaborate procedures, including the quantification of the systematic bias and the quantification of the random error via the mean absolute error or mean absolute percentage error. According to the law of large number and earlier research indicating that relative indices such as correlation coefficients necessitate a minimum sample size to stabilize, it was hypothesized that reliability indices follow an optimal sample size trend. In accordance with previous studies in correlation coefficients, this study highlights the importance of including high numbers of participants to receive stable reliability measures. The random error was not significantly affected by increased samples while providing important information about the performed standardization success in the testing, the study also underlines the relevance of reporting not only ICC-based reliability statistics but also the quantification of random errors.

KW - intraclass correlation coefficient

KW - law of large numbers

KW - measurement errors

KW - reliability

KW - repeatability

KW - Psychology

UR - http://www.scopus.com/inward/record.url?scp=105003144436&partnerID=8YFLogxK

U2 - 10.1080/1091367X.2025.2494998

DO - 10.1080/1091367X.2025.2494998

M3 - Journal articles

AN - SCOPUS:105003144436

JO - Measurement in Physical Education and Exercise Science

JF - Measurement in Physical Education and Exercise Science

SN - 1091-367X

ER -