Dendreon Corp fka DNDNQ

[exwannabe]

Selection bias in manufacturing.

Suppose we have a plant that makes 10mm ball bearings, but for whatever reasons they actually pop out of the machine at either 10.2mm or 9.8mm. Since we can't use the small ones, we sort these through a machine to discard those under 10mm, but this sorter is acurate to within +- .3mm (no bias in +/-). As we do this we record the measured size (from the sorter).

Now, take some of these bearings and measure them very precisly. They will be smaller, on average, than the measured size (since the impropoerly included bearings are all smaller than recorded, and the improperly excluded were all larger than recorded). This is straight math here without any placebos, trial size, of whatever.

It's really hard to argue against Dew's point that this bias exists. How significant in the case of Prov. is a much more difficult question.