Re: [Phys-L] Model Selection Statistics

[John Denker <jsd@av8n.com>]

On 4/15/19 11:56 AM, Paul Nord wrote:

> We're trying to
> study production asymmetry as a function of polarization which may depend
> on beam energy.

If you don't mind a question from the keen-grasp-of-the-obvious
department, based on no knowledge of the experimental constraints
or objectives.....

If you care primarily about the asymmetry, is there any chance
of doing a self-nulling differential measurement?  That is, take
the x-polarization data and the y-polarization data and subtract.
Under favorable conditions, that subtracts out the baseline.
That gets rid of all (or most) of the baseline, no matter how
ugly and complicated it is, with zero adjustable parameters.

That's not always possible, and there are eleventeen things
that could go wrong -- e.g. if the baseline conditions change
faster than you can change polarization -- but if everything
goes right it produces stunningly beautiful data.

(-: I've seen it happen! :-)

This is similar to the reasoning that led to the invention of
the lockin amplifier.  Lockins are used in some spectrometers
and in a bazillion other applications.  It's something physicists
know about but most other people don't.  (Roughly similar devices
can be found in the advanced engineering lab, but they are
considered super-advanced and arcane.)