Here's an example of why you have to be careful when using statistical measures without paying attention to the causal mechanisms.
Some people have a large waist size not because they are built big, but because they are obese. In the case of an obese person, some of the penis length is hidden due to the extra thickness of the fat pad, and a shorter measurement results. (Indeed, I have met grossly obese men with a negative penis length-- the organ was buried in a depression in their fat.) If there are many obese people in the population (as in America today), a negative correlation of waist size and measured penis size could result in the overall population statistics, even if there is actually a weak positive correlation between penis length and overall skeletal size.