What is the optimal sample size to perform areliability study within the motion analysis field ?

Powercalculation on experimental errors by Schwartz [3] ? We previouslyperformed a reproducibility and repeatability study of a foot segmentationmodel. During validation activities and review of the literature, we concludedthat reliability of kinematic data can be tested by several statistical waysand it seems that no agreement about the best method exist. In summary: · Coefficients of multiple correlation(CMC) is a measure of similarity of waveforms as gait curves. However, artificiallyhigh CMC values are related to large ROM and sampling rate, high number of testsubjects and neutralization of individual differences by natural offset removal[1, 2]. Roislien et al. concluded that CMC was unsuitable for assessing therepeatability and reproducibility of kinematic gait data.· Interclasscorrelation (ICC) does not take into account that gait curves are not aselection of points or summary statistics but continuous functions of time [2].Moreover,ICC depends on the variability inside the population studied. · So, experimental errors (EE) [3] wasselected as the statistical approach for assess the gait curves reliabilitylike in the studies of Caravaggi [4] and Deschamps [5]. We would liketo determine if our sample size (6 subjects and 4 raters) was large enough.In our study, powercalculation was not realized because we did not use statistical approach basedon p-value, and alpha and beta errors (P = 1 - beta). So, our questions are:1. How to calculate the power on experimental errors?2. Which statistical approach can demonstrate thatsample size was large enough?Thank foryour answers,PatrickSalvia and Céline Mahieu[1] Growney E, Meglan D, Johnson M,Cahalan T, An K. Repeated measures of adult normal walking using a videaotracking system.

[2] Roislien J, Skare O, Opheim A,Rennie L. Evaluating the properties of the coefficient of multiple correlationfor kinematic gait data.

[3] Schwartz M, Trost J, Wervey R.Measurement and management of errors in quantitative gait data.

[4] Caravaggi P, Benedetti M, BertiL, Leardini A. Repeatability of a multi-segment foot protocol in adultsubjects.

[5] Deschamps K, Staes F, BruyninckxH, Busschots E, Jaspers E, Atre A, et al. Reapetability in the assessment ofmulti-segment foot kinematics.

Powercalculation on experimental errors by Schwartz [3] ? We previouslyperformed a reproducibility and repeatability study of a foot segmentationmodel. During validation activities and review of the literature, we concludedthat reliability of kinematic data can be tested by several statistical waysand it seems that no agreement about the best method exist. In summary: · Coefficients of multiple correlation(CMC) is a measure of similarity of waveforms as gait curves. However, artificiallyhigh CMC values are related to large ROM and sampling rate, high number of testsubjects and neutralization of individual differences by natural offset removal[1, 2]. Roislien et al. concluded that CMC was unsuitable for assessing therepeatability and reproducibility of kinematic gait data.· Interclasscorrelation (ICC) does not take into account that gait curves are not aselection of points or summary statistics but continuous functions of time [2].Moreover,ICC depends on the variability inside the population studied. · So, experimental errors (EE) [3] wasselected as the statistical approach for assess the gait curves reliabilitylike in the studies of Caravaggi [4] and Deschamps [5]. We would liketo determine if our sample size (6 subjects and 4 raters) was large enough.In our study, powercalculation was not realized because we did not use statistical approach basedon p-value, and alpha and beta errors (P = 1 - beta). So, our questions are:1. How to calculate the power on experimental errors?2. Which statistical approach can demonstrate thatsample size was large enough?Thank foryour answers,PatrickSalvia and Céline Mahieu[1] Growney E, Meglan D, Johnson M,Cahalan T, An K. Repeated measures of adult normal walking using a videaotracking system.

*Gait and Posture*1997;*6*:147-62.[2] Roislien J, Skare O, Opheim A,Rennie L. Evaluating the properties of the coefficient of multiple correlationfor kinematic gait data.

*Journal ofBiomechanics*2012;*45*:2014-18.[3] Schwartz M, Trost J, Wervey R.Measurement and management of errors in quantitative gait data.

*Gait and Posture*2004;*20*:196-203.[4] Caravaggi P, Benedetti M, BertiL, Leardini A. Repeatability of a multi-segment foot protocol in adultsubjects.

*Gait and Posture*2011;*33 :*133-35.[5] Deschamps K, Staes F, BruyninckxH, Busschots E, Jaspers E, Atre A, et al. Reapetability in the assessment ofmulti-segment foot kinematics.

*Gaitand Posture*2012;*35*:255-60.
## Comment