About validity of conclusions based on multiple linear regression: a commentary on Kupelian et al. (2010).

摘要

Kupelian et al. (2010) concluded that "Results of [their multiple regression analyses of cross-sectional data] demonstrate the contribution of modifiable lifestyle factors to prevalence of ED [Erectile Dysfunction]" (Abstract, Conclusions-although of secondary importance, it should be made clear that the concept of prevalence, e.g., Mosby's Medical Dictionary (2009), is not equivalent to that of explained variance). They underlined "the importance of intervention studies targeting lifestyle changes, such as increased physical activity and smoking cessation, in the effort to prevent development of ED" (p. 24). We question the appropriateness of their methodology for scientific justification of such intervention studies. The preventive interpretation of Kupelian et al.'s analyses rests on an implicit nested structure of two main questionable assumptions. Resolving the validity problems linked to the first assumption only leads to the discovery of yet another problem.