Hi jesse, et al.
Nothing like statistics to ignite the passion in souls :)
While I recognize the concerns raised, I stand by the comments: the subset analysis of the KS study is statistically tenuous at best. This doesn't mean the ATIII without heparin group's findings aren't intriguing (as I wrote), or potentially valid (random chance not withstanding), or ones that would make me very happy (I own more shares than any member of the BOD but Cox) -- all of these are true -- they just aren't statistically meaningful.
People spend money on a lot of things, but personally, I will not design my own clinical trials, nor venture my own funds on the basis of data that are not statistically valid -- without other compelling arguments to prompt me on in the face of this deficit.
DNDN is only the latest to think statistics are just a quibble -- but at the end of the day, stats are where the rubber meets the road.
This said, I'll very much agree that "noticing relationships where others failed to see them" is the basis of good and great science -- it just isn't statistically valid 'proof'. So after Einstein made his predictions, a prospective trial had to wait some years (just wait for the eclipse . . .) before anyone (appropriately) cared. The same could be said of Fleming's discovery.
Data mining to help generate new hypothesizes on how to design new trials is great -- hats off to GTCB et al in doing so -- but my interns will tell you (as will Richard Pazdur) that it 'proves' nothing. They aren't wrong.
Best,
MTB
Market Data 
Markets 
AI
