Problems with Chi^2 calculations

[Michael Hansen]

Hi,

I have no tried to run several models of my modifications of Marceau Limousins A1703 model and I always get 0.000 in Chi^2. When I look in the bayes.dat, there are other values, but they are very low and there are very few of them and 0.000 dominates. This problem arises regardless of whether i am using my own models or the original model from Limousin.

This has changed from v. 6.7 to 6.8.

I have added a Histogram plot of Likelihood vs. Chi^2

Cheers,
Michael Hansen

[Michael Hansen]

Further information:

I tried to run the example_gnfw using the same settings on two different machines, using different OS, where all except (inverse 3 0.05) and (forme -1) has been left as they where originally.

On my laptop I get Chi2 = 0.0079 and on the University server I get Chi2 = 0.0035

I have also tried to run another model and compared it with the former 6.7 results and the Chi2 was almost the same.

So could the problem be a normbre too low? Can it be set above 128?

Cheers

[Eric Jullo]

Hi 

There might be a mismatch between relative and absolute coordinates for the multiple images catalog or the cluster member catalog.

Could you send the log message from lenstool. Does it look all fine? 

Cheers

Eric

--
--
_____________________________________________________________________
LENSTOOL: http://projets.lam.fr/projects/lenstool/wiki
To post to this group, send email to lens...@googlegroups.com
To unsubscribe from this group, send email to lenstool-u...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/lenstool?
---
You received this message because you are subscribed to the Google Groups "lenstool" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lenstool+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Michael Hansen]

Hi Eric,

Do you mean the verbose message? I have included it in this message. As far as I can see, there are no problems.

Cheers,
Michael

[Eric Jullo]

Hi Michael,

You're right. Everything looks fine. 

If you try with source plane optimization, do you get the same chi2 values? 

You're saying this appended between v6.7 and v6.8?

Cheers

Eric

<lenstool_verb.log>

[Michael Hansen]

Hi Eric,

If I try with SP opt. I get a Chi2 of ~400. And yes, the difference has been since v. 6.8. But it gets even stranger. I have tried running a model where I changed some minor parameters (added another NFW profile) and it appears to give the correct result, even though it is a complete copy of another parameterfile. But I have to wait for the final result.

Cheers,
Michael

[Michael Hansen]

Hi Eric,

I've got some new info. This seems very strange!
COMP: Cm: multiplicity: 2 at (41935786167992024402231296.000,-220490868290036229354815488.000): found 0 image(s)  
COMP: Cl: multiplicity: 2 at (36783010249886441012199424.000,-220332964287768970364190720.000): found 0 image(s)
COMP: Ck: multiplicity: 2 at (40793953194226496051347456.000,-217099919730202231372251136.000): found 0 image(s)
COMP: Cj: multiplicity: 2 at (67349327426221686243983360.000,-238742690739200567350919168.000): found 0 image(s)
COMP: Ci: multiplicity: 2 at (-110106907934467049377497088.000,-214058668071802877526409216.000): found 0 image(s)
COMP: Ch: multiplicity: 2 at (-124110036459946797107773440.000,-208117800910607847910277120.000): found 0 image(s)  
COMP: Cg: multiplicity: 2 at (-185910998318626833495490560.000,-187129917867824185180422144.000): found 0 image(s)
COMP: Cf: multiplicity: 2 at (-221269721565992949095858176.000,-100793680473899466714251264.000): found 0 image(s)
COMP: Ce: multiplicity: 2 at (-222913634841946608839426048.000,-93578715279666019590733824.000): found 0 image(s)
COMP: Cd: multiplicity: 2 at (-226382156138701895757725696.000,-95612220233191303076118528.000): found 0 image(s) 
COMP: Cc: multiplicity: 2 at (-205724255642768315963670528.000,-112165564288532004365926400.000): found 0 image(s)
COMP: Cb: multiplicity: 2 at (-206230639536411709592829952.000,-110737284666882978838740992.000): found 0 image(s)
COMP: Ca: multiplicity: 2 at (-208646783658904712969715712.000,-106776174362776906875535360.000): found 0 image(s)
COMP: 23.1: multiplicity: 2 at (-298467344613826162635833344.000,57627497731872962457370624.000): found 0 image(s)
COMP: 21.1: multiplicity: 2 at (-57932840000380720537141248.000,-141542280857932077870350336.000): found 0 image(s) 
COMP: 20.1: multiplicity: 2 at (-33125675243966890650894336.000,-146833558151086065680646144.000): found 0 image(s)
COMP: grid at z=0.889 (0.630)
COMP: 1.5: multiplicity: 2 at (-387366711885233559437312000.000,-150211445202986146748956672.000): found 0 image(s)
COMP: grid at z=1.123 (0.688)
COMP: 14.4: multiplicity: 3 at (-272717123226197042455904256.000,-1206412759264383016766537728.000): found 0 image(s)
COMP: grid at z=1.259 (0.712)
COMP: 13.4: multiplicity: 2 at (-410628425302829782284632064.000,-367960574327588296924856320.000): found 0 image(s)

At least 1.5 is one of the multiple images going into the model af constraints.

Cheers,
Michael

[Kneib Jean-Paul]

strange - looks like a coordinate transformation problem or some memory issue - can you post your parameter file?

Jp

[Eric Jullo]

Hi Michael,

Did you put add the line #REFERENCE, in the header of your multiple images file? I remind you the syntax

If your catalog is in arcsec coordinates relative to a given position RA_REF, DEC_REF then you should add the line

#REFERENCE 3 RA_REF DEC_REF

If your catalog is in absolute degrees coordinates, then you should add the line

#REFERENCE 0

Cheers

Eric

[Michael Hansen]

Hi,

I have tried to insert the #REFERENCE line in both the multiple image file and the perturbing galaxy file, in the original Limousin model. It didn't work. I am still running my own model, which I have posted the relevant files for here.

Cheers

[Michael Hansen]

Hi all,

I might have some sort of solution to this problem.

What I have noticed is that all the galaxies giving me problems where galaxies with a magnitude higher that what was given to Lenstool in the as the mag0 reference. I have now tried to test my model setting the mag0 to the lowest magnitude in my catalogi.

Cheers,
Michael