[yt-users] Fitting an absorption spectrum

Wed Mar 5 09:02:22 PST 2014

Hillary & Pengfei,

I'm going to hopefully add to this discussion. I've been working for a
little bit on some improvements to the fitting procedure, primarily because
I've been interested in being able to fit spectra with specified
uncertainties and a significant amount of noise.

Although down the line my (quite extensive) modifications to the fitting
procedure may be problematic to merge with the main repo, I would be very
interested in obtaining a new test case to try out my version on. Even
though my modifications were in the context of noisy spectra, they still
may perform well in dealing with the strange results you are seeing.

I may or may not be of help here, but if you don't mind sending your
absorption spectra over to me to test it out I would appreciate it. The
turnaround time on this may be long, but I should be able to get to play
around with it soon and let you know what happens.

If things go well, I'll point you to my fork where you can obtain the
modifications I've made.

Andrew E.

On Wed, Mar 5, 2014 at 10:30 AM, Hilary Egan <hilaryye at gmail.com> wrote:

> Hi Pengfei,
>
> I'm glad at least some of my fixes worked! I'm not sure whats going on
> with your b-value distributions. I've performed the same analysis for my
> absorption spectra and I dont see anything like the peaks at ~40 km/s that
> you do (see attached). At the high end, I do see a very slight excess of
> absorbers at the high end, but again nothing like what you're showing.
>
> This leads me to the conclusion that the fitting process works well for
> spectra generated like mine, but behaves somewhat poorly for yours. Having
> looked briefly at one example lightray of yours (offlist) I can definitely
> see that yours looks qualitatively different than mine and has a lot more
> lines and those lines are much narrower, although having little idea what
> exactly you're looking at I can't really extrapolate beyond that.
>
> Going forward I can relatively quickly add something that throws away
> lines with these parameters as they're clearly unphysical; however, I'm
> somewhat uncomfortable doing that as its probably an indication of a larger
> problem for a certain class of spectra. I wish I could guarantee that I
> could look at this and completely fix it for you, but I really don't have
> more time to devote to this project at the moment. If you're interested in
> helping look at it I'd be more than happy to point you in the right
> direction and we can chat offlist/in the irc. Otherwise I'll update the
> pull request to just remove those lines, and heavily comment it so any
> future user can remove/play around with it if they're unsatisfied with this
> stop-gap measure.
>
> -Hilary
>
>
> On Tue, Mar 4, 2014 at 11:32 PM, Pengfei Chen <madcpf at gmail.com> wrote:
>
>> Hi Hilary,
>>
>> Thank you very much for your update and I'm sorry for the late reply.
>> I've tested the new code with two datasets. The results are attached. The
>> abnormal peaks in the statistics of N have disappeared, and the abnormal
>> peaks for b are much lower now, but still visible.
>>
>> Thank you very much for your effort in fixing this! Please tell me if you
>> need any test data from me.
>>
>> Pengfei
>>
>>
>> On Fri, Feb 14, 2014 at 3:30 PM, Pengfei Chen <madcpf at gmail.com> wrote:
>>
>>> Hi Hilary,
>>>
>>> Thank you! I'm looking forward to it.
>>>
>>> Pengfei
>>>
>>>
>>> On Thu, Feb 13, 2014 at 7:32 PM, Hilary Egan <hilaryye at gmail.com> wrote:
>>>
>>>> Hi Pengfei,
>>>>
>>>> I have a fix for this issue that just needs a little more testing! I'll
>>>> push it shortly and reply to this message when it's been accepted.
>>>>
>>>> -Hilary
>>>>
>>>>
>>>> On Thu, Feb 13, 2014 at 8:04 PM, Pengfei Chen <madcpf at gmail.com> wrote:
>>>>
>>>>> Hi all,
>>>>>
>>>>> I've used the
>>>>>
>>>>> yt.analysis_modules.absorption_spectrum.absorption_spectrum_fit
>>>>>
>>>>> script to fit an Lya absorption spectrum generated by yt.
>>>>>
>>>>> But there are some abnormal peaks when I did statistics of line width
>>>>> b and column density N. Two figures are attached. I've tried with different
>>>>> redshifts and all of them had peaks in b-distribution at init_b (which is
>>>>> 40km/s here) and maxb(which is 300km/s here). For the N-distribution I got
>>>>> peaks around 10^12 cm^-2 at every redshift. The parameters I used were:
>>>>>
>>>>> HI_parameters = {'name':'HI',
>>>>>                  'f': [.4164],
>>>>>                  'Gamma':[6.265E8],
>>>>>                  'wavelength':[1215.67],
>>>>>                  'numLines':1,
>>>>>                  'maxN': 1E22, 'minN':1E11,
>>>>>                  'maxb': 300, 'minb':1,
>>>>>                  'maxz': 5, 'minz':0,
>>>>>                  'init_b':40,
>>>>>                  'init_N':1E14}
>>>>>
>>>>> Could someone please tell me how to solve this? All the analysis were
>>>>> done with the latest development version of yt and I made no changes to the
>>>>> source code.
>>>>>
>>>>> Please tell me if you need any datasets.
>>>>>
>>>>> Thank you!
>>>>> Pengfei
>>>>>
>>>>> _______________________________________________
>>>>> yt-users mailing list
>>>>> yt-users at lists.spacepope.org
>>>>> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>>>>>
>>>>>
>>>>
>>>> _______________________________________________
>>>> yt-users mailing list
>>>> yt-users at lists.spacepope.org
>>>> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>>>>
>>>>
>>>
>>
>> _______________________________________________
>> yt-users mailing list
>> yt-users at lists.spacepope.org
>> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>>
>>
>
> _______________________________________________
> yt-users mailing list
> yt-users at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>
>

-- 
Graduate Student
Columbia University
Department of Astronomy
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.spacepope.org/pipermail/yt-users-spacepope.org/attachments/20140305/8a3755e6/attachment.htm>