RUU #11: Strong evidence + physics

This clip contains thoughts about physics and predictions from physics as examples of what's meant by 'strong evidence'. Two cases are studied, namely the prediction and subsequent discovery of Neptune and the predicted magnetic moment of the electron. The point is to illustrate how much a single piece of evidence can mean for two different models. the models are a somewhat archetypal, there are a lot of middle-stances possible as well as other possibilities. What I want to study here in basically a how established physics works compared to less accurate stances based solely on similar observations.

The first example is taken from E. T. Jaynes book, probability theory - the logic of science. Jaynes was originally a physicist who later developed ideas in the intersection between physics and statistics. (While my contributions to either fields are extremely small, I am in the same boat in that matter.)

Pictures are fetched from wikipedia articles.

Category:

Howto & Style

Tags:

• Bayes
• Bayesian
• probability
• probabilistic
• reasoning
• physics
• strong
• evidence
• Newton
• Neptune
• magnetic
• moment

[U83r1337hax]

lol what the fuck you guys talking about this shit is boring get a life and or die :D

[BrunoTheQuestionable]

Sounds good.

I'm still trying to work out how we can be moving relative to the Cosmic Microwave Background !

[trondreitan]

Indeed. It could be that I could use the Michelson-Morely experiment (1887) in order to check the then extraordianry claim that either Earth wasn't moving relative to the aether or there's no aether at all. The data seems to be available. That must wait, though, since I haven't covered continuous distributions, yet.

[BrunoTheQuestionable]

Yes indeed. You would think such crucial data would be more widely available.

[trondreitan]

Perhaps, but seem to remember that these spectral lines were predicted by Bohr's semi-classical models too. Still, that was an earlier step in the quantum direction. The problem is still to get hold of the data and the data uncertainties.

[BrunoTheQuestionable]

I quite like this example - quantum theory is so weird that it may have been the most difficult revolution in the history of physics. The data and the theory gradually developed over more than 50 years. Was there a turning point where it could no longer be denied ? Perhaps the prediction of the spectral lines of hydrogen by the Schrödinger equation ?

[trondreitan]

Maybe I should think about a completely different example. For isntance, a tweaked version of the medical example from clip 3 and 10 with a very rare disease, like smallpox or the bubonic plague (rare now). Or perhaps a murder mystery with some extraordinary claim (the butler didn't do it).

[trondreitan]

Well I assigned 50% probability before the ten digit precision magnetic moment data, but more to illustrate the impact of that particular data than anything else. Quantum field theory was really well established in advance. But there may be some instances earlier in the developement of quantum theory that may be used as an illustration. The trouble is to get a fix on the measurement precision in the experiments that led to quantum theory.

[BrunoTheQuestionable]

Yes, relativity is not a great example. You're Quantum Theory example is more data driven - you assigned an a priori probability of 0.5 but in the absense of any data (i.e. not even any knowledge of the photoelectric effect), Quantum Theory would seem ridiculous. Its necessary to make clear the starting point.

[trondreitan]

Actually, I think relativity is an extreme example, as Einstein's reasoning around the principle of invaraince and how that applies to Maxwell's equations and the Newtonic problems you mentioned could very well be said to have increased the prior probability for relativity to plausible even before data. Normally model-building is more driven by data and less by basic principles, in which case you would need strong evidence in order to form a new model.