Summary of chop-nod analysis telecon of Oct 29, 2008

Mike, Giles, Lero, Hiroko

_____________________________________________________________

Lero's analysis of M 82

Note: a document was posted by Lero last week.

First, Lero gave some background:

we have 55 good files on M82.

they are now organized into 5 bins of 11 files each

nominally we have detections, some of 4-sigma significance, when all 55 
are combined

but Lero's chi2 analysis implies we have to inflate errors, in which 
case we would not get 3 sigma detections

the idea is to look again to try to establish reliability in a different 
way.

referring to the first page of memo (p 3), this explores varying the 
smoothing

Lero likes top right best (9.5, 9.5 smoothing)

2nd page of memo (p 4);: explore the 2 sigma vectors

3rd page of memo (p 5): divide entire data set into two equal bins; 
first bin shows no vectors; perhaps tau*sec(z) was lower for later files

plan is to next try "equal noise" bins; bins with similar sharpcombine 
errors in Q and U are preferred.

Giles notes that we can get this about right by combining the first two 
bins.  Then we will have four bins: A=12, B=3, C=4, D=5

Giles points out that the bottom left of fig 4 and left of fig 5 would 
represent bins A+B and C+D respectively.  They do not confirm one another 
but they do not disagree.

Hiroko asks: how was the data editing done?

Lero: this came from distribution of noisy pixels (masked out) with 
respect to the source.  tau was not used as a criterion for rejecting 
files, since source only was observed in good tau.
 
We discussed what to do next.  We agree that the 4 bins (A B C D) are 
the way to go.  Then the idea is to compute the reduced chi square from 
four and two (AB vs. CD) bins.  Then look at the chi2 maps, and decide 
how to inflate.

Lero asked about the problem John discovered with sharpinteg.  We think 
Tristan has a routine to check for this bug.  Lero wants the routine 
when its ready for distribution.

_____________________________________________________________

Mike's work:

Mike has new results for 1333 4A and 4B, and a new way to estimate the 
error inflation factor that is coded up into chi2.

Giles is happy about the former, and curious to learn more about the 
latter

We will discuss all this more in future

_____________________________________________________________