Sunday, February 26, 2012
Satan loves stats

Paul Homewood in UK discusses the funneling of most remaining science funding toward Green shit. This is predictable; the Green tumor knows how to claim all the blood supply of its victim until the victim is completely dead.

What really caught my attention was this sentence:
As a result certain areas of science are having all funding withdrawn, for, at the least, the next year. For instance PhD fellowships will no longer be available for Engineering graduates, and will be limited to only Statistics & Applied Probability for Mathematicians, as the EPSRC website makes clear.

The agency knows perfectly goddamn well what it's doing. Preferring stats over engineering isn't as obvious as funding projects with "Climate Change" in their titles, but it's actually worse in the long run.

Engineering is a constructive use of math.

Statistics is a destructive use of math.

Satan loves stats.

Stats serves the climate pseudoscientists. With the help of stats they are able to make a completely false theory give very small positive results. Without it, they'd have to look directly at the data and see the complete lack of connection between temperature and CO2.

Stats serves the sociologists who provide a "factual" basis for court decisions on Die-Versity, smashing great cities into rubble. Without it, they'd have to acknowledge the obvious factual differences between living societies like Korea and dying societies like Britain and US.

Stats serves the Wall Street Mafia and their slaves in CongressMart. With the help of stats they can find the most likely bet (derivative, option, etc), and then kill companies directly, or design laws to kill the proper set of companies to guarantee the bet.

Stats serves political operatives like Karl Rove and Dick Morris. With stats they can find the best possible gerrymander to guarantee My Party Wins Forever. With stats they can locate the 100 voters who really count in a national election, then figure out how to spend a trillion dollars of everyone else's money to buy those 100 voters. (See Medicare Part D.)

= = = = =

Here's reality: A difference that's worth acting on will appear clearly in the properly measured data of a properly designed experiment or observation. You won't have to tease it out with stats. If you can't design a good experiment to pin down the relevant variable, or locate a 'natural experiment' that separates the variables, you don't have a point. Give up and take a nap.

If your model "requires" thousands of variables, you have ALREADY ADMITTED that you are not objectively analyzing real data. You have ALREADY ADMITTED that you are seeking a predetermined result which will advance your career, improve your status, and enrich your grant account.

Medical research tacitly recognizes this reality. If Drug A is curing most of the terminal cancer patients, the clinical trial stops and all the cancer patients get Drug A immediately. No need for stats. If Drug B is killing some patients, the trial stops and Drug B is discarded forever. No need for stats.

If I were emperor I'd ban statistics entirely, except for the median. Since I'm not emperor, I can only show the problems and wish for lower subsidies to stats and higher subsidies to engineering and agriculture.

Labels: ,