Mixed
Absolute variance statistics (SDir, lnVR) are insufficient for detecting true interindividual response variation because they are confounded by mean-variance relationships (where higher means lead to higher variances); the Log Ratio of Coefficient of Variation (lnCVR) is required to isolate true response heterogeneity.
When evaluating resistance training results, do not assume that a wider range of outcomes (some people growing a lot, some a little) is solely due to individual 'trainability.' It may simply be that those who grew more had higher starting strengths, which naturally scale with greater variance. To truly understand why some people don't respond, you must look at relative variation (coefficient of variation) rather than absolute changes. For the individual, this means that 'average' program prescriptions will inevitably fail some people, not because the program is bad, but because biological scaling laws dictate that larger gains come with larger variance.
As we have seen from the SMD and lnRR models, RT interventions increase mean scores. Thus, if there is a mean-variance relationship in the data, an increase in the mean alone may be fully responsible for any apparent increase in variation. As such, we cannot rely solely on absolute comparisons of variance such as the SDir and lnV R to determine whether interindividual response variation is actually present. The lnCV R can be used to overcome this issue
Why this rating
Theoretical/statistical derivation and re-analysis of existing data.
Source
Meta-analysis of variation in sport and exercise science: Examples of application within resistance training research
James Steele et al. · Journal of Sports Sciences · 2023
This is one finding among thousands. Every one is graded and traced to its source, so you can see what the evidence actually supports. Browse the research →