>The positive connotation of data mining is examining ...data from various angles to form a hypothesis for further testing.
In other words, science is good (and retrospective analysis can be good science).
My stats prof Andy Barron gave me a copy of his dissertation. I almost dropped it on the floor of his office (it was only 3/16" thick). He proved that systematic data mining sufficed to obtain (in the fullness of time) all truth: if the world is lawful the candidate laws on formulation are recursively enumerable and the candidate data sets against which they may be tested likewise so diagonalization yields a method which converges to the desiderata.
>>Data mining has both a positive and a negative connotation. The negative connotation applies when a pivotal trial fails and the sponsor looks around for a way to salvage the clinical program without having to run a new trial. This is what a great many biotech companies do all the time. It hardly ever fools the FDA but it sometimes fools investors long enough to prevent a total collapse of the stock price.<<
That is exacly what ABT did after the FIELD trial failed its primary endpoint and nobody bought it. If you don't hit the primary endpoint then nothing in the secondary stuff is really valid as I see it. Then ABT bought KOSP to get Niaspan, among other things. Jaharis sold KOSP when it looked like Torcetrapib was going to take over the HDL market. I think ABT knew something Jaharis didn't. Now ABT is promoting Niaspan over Tricor and touting outcomes data from HATS and the ARBITER studies.
Market Data 
Markets 
AI
