On Sun, Dec 22, 2013 at 8:41 PM, Edmund Storms <[email protected]>wrote:
> Yes, if 4 don't work, try 5. If 5 fail, try 50. As they say, with enough > variables, all data can be fit. It never occurs to these people that their > basic model may be wrong. They just keep adding variables and complexity > until all the points fall on a line. > > Ed Storms > I always wanted to know what theoretical physicists have to say about that argument. I just saw in the swedish television a nice program that was a filosofical discussion about science in general, and that question came out. The feeling of the expert was that it was too much things that fell on place that they argue that the theory should be pretty much correct and that the theory predicted things that was shown correct. But then when I take my statistical and mathematical hat on I just see to much opportunity to add variables that are fitted in experiments and then the whole thing matches a theory of a certain order, that should by some expansion fit to a more conventional albeit perhaps a bit more nonlinear model. Then of cause the high dimensional theory would predict the right things if what is probed is for example the second order effects of the model. Another pet peeve of mine is the popular argument that because we live it must be true argument, which is fine if one also supply alternative arguments. As an example take the realization that if the physical constants was just a little of, then the universe would not allow us to live and because we live, the parameters are as they are assuming a multiverse. Another argument is that the model of all things have really fewer parameters and hence that the parameters really are dependant. But you never hear that possibility in the TV shows, which shows a good piece if overconfidence of what we have today. Cheers! /Stefan
