Comparative analyses of a variable measured twice or against a "gold standard" technique should explore the existence of any fixed and proportional biases between the 2 measurements. Levels of agreement (LOA) consider these biases together and least products regression (LPR) consider their effect independently. To compare the use of LOA and LPR, the peak torque extension (PTE) of the knee at 2.1 rad/sec during isokinetic dynamometry was obtained on 2 separate days (N = 17). The mean PTE (with standard deviations in parentheses) was found to be 93.6 (13.9) N · m on Day 1 and 92.5 (11.5) N · m on Day 2. The LOA were 1.06 ± 10.80 N · m (95% confidence), and the LPR's (with 95% confidence intervals in parentheses) intercept was -17.7 N · m (-37.4 to 2.03) and slope was 1.20 (1.01 to 1.40). LOA and LPR are suitable techniques to compare 2 measurements and, because the levels are large and the slope does not encompass 1, suggest that the knee's PTE at 2.1 rad/sec is unreliable.
[1]
J. Ludbrook.
SPECIAL ARTICLE COMPARING METHODS OF MEASUREMENT
,
1997
.
[2]
D. Altman,et al.
STATISTICAL METHODS FOR ASSESSING AGREEMENT BETWEEN TWO METHODS OF CLINICAL MEASUREMENT
,
1986,
The Lancet.
[3]
W. Zhu,et al.
Making bootstrap statistical inferences: a tutorial.
,
1997,
Research quarterly for exercise and sport.
[4]
D G Altman,et al.
Comparing two methods of clinical measurement: a personal history.
,
1995,
International journal of epidemiology.
[5]
Margarita Pérez,et al.
Saliva Composition and Exercise
,
1998,
Sports medicine.