In consistent fit results when using "LineRatioFit"

[Carl Ferkinhoff]

We are trying to use PDR Toobox to model the PDR of high-z galaxies using observations of the galaxies' OI145, CII, and FIR. However we've run into an issue when use "LineRatioFit". With these observables there often to minimums in chi-squared in the G and n plan, with a lower G high n option often being unphysical. When we attempt to moddel the observables, "LineRatioFit" returns best fit values that fall neight on the high G, low n solution nor the low G, high n solution. But a combination of both. 

Here is an example. We start with the following measurements:

After running the line ratio fit with least squares method, we get the following results.

Graphing this results, we see that LineRatioFit as selected the density from the high G solution, and the radiation field form the low G solution . . . which doesn't follow on the observed mesurements!

If we use the emcee method we get the same result.  What is strange, is if we plot the chi-squared map, it correctly gives the best fit value as the high G, low n solution

Is there a way to limit the G and n range so LineRatioFit reports the correct best fit result?

[Marc W. Pound]

hi Carl,

The general issue of two nearly equally plausible fits has come up before (we even point it out in our 2023 paper), and being able to clip the model space to remove unphysical solutions would be a reasonable workaround.  There is no API for this but it still may be possible.  I am surprised that emcee gave the same strange result since that uses a completely different fitting package.

- What version of pdrtpy are you using? 

- Please send me your full notebook or script.

- Could you also report this as a github issue? Helps me track problems better.  https://github.com/mpound/pdrtpy/issues

best,

Marc

--
You received this message because you are subscribed to the Google Groups "PDR Toolbox" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pdrt+uns...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/pdrt/9099fb9e-d095-444c-a512-c239d5a0d668n%40googlegroups.com.

--

Dr. Marc Pound
Research Scientist
Astronomy Department
University of Maryland, College Park

https://www.astro.umd.edu/~mpound

mpo...@umd.edu | 301.405.1520 | https://umd.zoom.us/my/marc.pound

[Carl Ferkinhoff]

Hi Marc,

Thanks for a quick reply! Yes I was suprised the issue showed up in the emcee method as well. I'll send your the further details and submit it as an issue to github . . . but it will probably be a couple days. My main questions was if there was an API method to limit the phase space.

For versions, it happens in the current version v2.6.4 and v2.5.4

Carl