News for hackers. Stuff that matters.
Re:The press-release conviction of a biotech CEO and its imp... by PeterisP on Thursday October 03, 2013 @03:01AM (#6487547)
If you divide patients in 20+ arbitrary groups (say, based on first letter of their name) and give them a 0 efficiency treatment, then almost always you'll see a group with a p-0.05 significant improvement; that' just basic stats.

So, if you afterwards discover some class of patients with this different response, it is grounds only to investigate that group further separately, not anything more - claiming that there is such an effect is fraud. You can't make extra claims until you have verification of that separate group.

For an illustration, see http://xkcd.com/882/

Re:The press-release conviction of a biotech CEO and its imp... by mbreese on Thursday October 03, 2013 @02:17PM (#6490828)
Of course, but they divided the patients into two groups: Mild/moderate and severe clinical phenotypes. It's not optimal, but it is far from subdividing into 20 classes or making up arbitrary classifiers until you got the number you were looking for. The clinical phenotypes were already established prior to the analysis.