Summary of chop-nod analysis telecon of July 8, 2009

Giles, Lero, John, Hiroko


_____________________________________________________________

Hiroko's analysis of IRAS 20126:

(see also July 1 posting on analysis logbook, posted after the last telecon)

Hiroko reported progress on the following action items from last week:

- check on weights - DONE, THEY ARE FINE

- check how cropped chi_2 depends on number of bins - SEEMS FINE (see the 7/1 posting); SHE USED A 
23 ARCSEC CIRCLE WHICH SEEMS GOOD

- debias the vectors with "/debias" if going much below 3 sigma - DONE

- consider the fact that i.p. is uncertain at 0.1% level - NOT A PROBLEM, P > .3%

problems that still need to be addressed, as of July 8:

- when seeing whether red_chi_2 changes with number of bins, we need to be sure that the sky region 
used to calculate red_chi_2 does not change when one changes the bin number.

- show more levels below 10 % (levels keyword)

Hiroko changed from skip 4 to skip 3 and the map changed a bit.  This raises the general issue of 
how do we deal with the freedom of choice that we have in deciding which vectors to show.  Recall 
that we can choose what fraction of vectors to show (skipv = 1, 2, 3, 4, etc) and also which 
specific vectors to show (i.e. where do we start counting; I forget the option name but it can be 
done).  Lero suggests just going with the default as a precaution to avoid being inadvertently 
biased.  Giles likes this.

John notes that debias option also has the effect of rejecting more vectors, i.e. it has the effect 
of a stricter cut.

_____________________________________________________________


Lero's analysis of M82:

- we discussed Lero's e-mail to sharp-software sent July 5th

- some discussion of whether the vectors that seemed too long before might actually be OK in the 
latest analysis.  Lero says we have 18% however, which after debiasing may still be uncomfortably 
large.

- Lero has contours of I/sigma-I that show a lowest contour at the value 0.1*peak which would seem 
to correspond to I/sigmaI = 4.  Some vectors lie well outside this contour so we may have 
justification for cutting some vectors.  We need to understand how sigma-I is calculated.

_____________________________________________________________