Send letter to journal:
Re: Validity comparison of different protocols

Dear Editor,

In the last years, validity and reliability of tests used is exercise science has gained a lot of attention. Since the review of Atkinson and Nevill (1998), many works has published addressing these matters.

Baltaci et al., in the present work, assessed the validity of three different forms of sit-and-reach (SR) test and compared the results with flexibility of hip joint for flexion using a goniometer. However, authors used the Pearson Correlation Coefficient (r) as a validity scale.

Since the work of Altman and Bland (1983), it is well know that r is not a good way to assess validity or reliability. Really, this notion was already shown before (Hallman and Teramo, 1981). The reasons, well described in the work of Altman and Bland (1983), can be resumed in two points. First, Pearson r is the relation between variation of values across individuals and variation within individuals. If variations between individuals are large, correlation will be large, independently of variation within individuals. In this case, variation between individuals is large, what can be demonstrated by large standard deviations in Table 1 of the article.

Second, if we want to use this approach to validity, we need to regress values of one method (in the case of different scales) and estimate de validity by the Standard Error of Estimate (SEE). The problem here is that SEE is a value that will change depending of the distance from the mean.

Authors also stated that SR and BSSR are highly related to hamstrings flexibility and the CSR test was not related. But, at conclusion, authors stated that all SR tests had similar criterion related validity. Based on the r coefficients, all three tests are significantly related and none had a strong relation. CSR presented a week correlation (0.22 and 0.21, comparing with GML, as the authors) and SR and BSSR had a moderate correlation (SR=0.63, BSSRL=0.37, BSSRR=0.50).