This self-selection tends to invalidate the study

Not to mention the fact that almost no treatment of a "study" in popular media distinguishes between correlation and causation. There is a difference, a fact lost on almost everyone these days.

[I lost .4 pounds last Thursday commenting on Hacker News. Therefore, typing causes weight loss :-) ]