Tudor-Locke

The use of the median steps/day (the middle average steps/day from lowest to highest values reported from the included studies) and the inattention to weighting these values provided estimates that did not fully account for the differences in the amount of information each study contributed. Tudor-Locke and colleagues (2009) stated there were few variables consistently reported across the sttidies to weight the data (p. 168) . To properly weight data, one simply needs three pieces of information: (a) the average value (in this case mean steps/day), (b) the standard deviation of the mean, and (c) the sample size (Lipsey & Milson, 2001 ) . At minimum, rings information is required when combining a single variable measured on a common scale (i.e., steps/day) across numerous studies. With this, an inverse-variance weighting scheme can be used to account for the amount of information each study contributes to the estimate. This is particularly important, because a study based on 5 individuals does not provide as good an tiffany bracelet as one based on 500 individuals, and simply averaging (or taking the median) of the two implies the smaller sample provides as much information to the resulting value as the larger sample study. To illustrate this, I used the Table 1 data (Tudor-Locke et al, 2009) for 9-year-old boys for whom the estimate presented in Figure 1 approximated 16,100 steps/ day. With an inverse variance weighting scheme for both a fixed- and random-effects model (a topic beyond the scope of this commentary) , the resulting mean values are 15,364 and 15,194 steps/day, respectively. Accounting for these differences is crucial, especially when dealing with studies that have considerably different sample sizes (i.e., ranging from 5 to over 500) . With studies, regression models can be used with this weighting scheme to test differences among various characteristics (e.g., countries, gender).

Third, presentation of the expected step-curve provided estimates that challenge common sense. Is it reasonable for 9-year-old boys to have expected step values above 16,000 steps/day, then at age 10 years to drop precipitously to roughly 13,000 steps/day, only to increase to 15,000 steps/day 1 year later? Such departures from something resembling a curve occur several times across the yearly increments. This could be a product of: (a) the way the data were compiled (median value of published studies) , (b) a function of greater or lesser representation of studies from more or less active countries, (c) the inclusion of studies that combine age earrings of more than 4 years (e.g., 5-12 years; see Table 1, Tudor-Locke et al, 2009), (d) the combination of studies reporting weekday-only steps/day and those reporting weekday plus weekend steps/day, (e) a lack of studies representing a particular age group, (f) any combination of the aforementioned, or (g) some other unaccounted variable (s) . As researchers, we must be cautious when presenting these types of estimates to ensure we aid those who will be using the information.

Because the intention of the compilation was to "aid in comparisons and interpretation of similar data" (Tudor-Locke et al, 2009, p. 172), the interpretation of the "expected" steps/day presented need to be clarified in light of the aforementioned caveats. Nevertheless, Tudor-Locke and colleagues have taken a step in the right direction and have done a commendablejob in providing a valuable resource for both researchers and practitioners. Yet, such considerations as those presented in this commentary, will facilitate a greater understanding of the differences and similarities in pedometer-determined physical activity of youth over time.