Originally Posted by
MoAlpha
Well, I like your worldview, but not that paper!
On a technical level, they computed way too many correlations with no attempt account for the number of statistical tests, greatly increasing the likelihood of finding spurious relationships. Bottom line, those p values are worthless. The other problem is that, as one might expect, many of those dimensions are “autocorrelated,” meaning that they are not independent of each other and they should not have been treated as such. The data set was begging for a principle component analysis, which reduces the dimensionality and finds the underlying, meaningful, trends. Sadly, however, the sample size of 50 is far too small for a valid study with this many dimensions. As a scientific journal editor, I would not have sent this out for review.
Even if we accept the analysis and conclusions, I don’t find it surprising or informative that parameters associated with physical robustness show correlations. Finally, I can’t believe there was no mention of smoking in a sample of middle aged Lebanese men!
Give me an interventional study in anyone, any day!
Sorry to dump like this, but, hey, it’s what the taxpayers pay me for.
Tell me what you really think,
The perils of reading and referencing papers from the NHI.