Summary of chop-nod analysis telecon of June 25, 2008

Mike, Hiroko, Giles, Lero, Tristan, Martin

_____________________________________________________________

Hiroko's analysis of IRAS 20126

see Hiroko's updated posting on analysis logbook:

- pointing: John had suggested that "by eye" inspections can help 
discriminate between method 1 and method 2

Hiroko found that the source is bright in all cases - see her posting - so 
she decided to use the fitgauss *unsmoothed* FAZO FZAO.

Giles found with L1527 that just because the peak looks good by eye does 
not mean that fitgauss will work well, and that one could compare the 
pixel where the peak appears to be "by eye" with the fitgauss result, by 
making the comparison in "pixel space".  Note that one cannot easily make 
such a comparison in FAZO/FZAO space, which is how Hiroko's smoothed FAZO 
FZAO is stored.  Giles noted that the cases where fitgauss fails are 
usually accompanied by wild fluctuations in FwHM (factors of 2-3).  
Hiroko kept an eye on this parameter and did not see such fluctuations.  
So, actually, we are probably fine with the pointing now.

Also note that the polarization results do not depend sensitively on the 
detailed method used for pointing correction, as the new results are 
similar to the ones from Hiroko's first memo.

- background subtraction: John had suggested looking at plots of bg 
levels, and perhaps throwing away outliers.  Hiroko has now done this, and 
the results don't change much.  Note she has posted the results of John's 
-idl option.

- Giles suggests adding polsharp4 options used to the document

- Giles suggest that "bins" should be sequential in time, ideally

- Hiroko presented her results from chi2.  The map-average is 4-5.  (This 
implies we have ~twice as much noise as what is included in our error 
bars.)

- The chi2 results for individual vectors are included in the memo, but 
they are not very meaningful as they are expected too show too much 
statistical variance since the chi2 is computed from just three bins for 
just one position.  Lero suggested having more than three bins.  Giles 
suggested binning groups of similar-looking vectors.  Either method, or 
both, should give meaningful info on the red.chi2 of the data, and how it 
might change over the map.  Another suggestion was to compute the chi2 
over just the region of the map where we have vectors (i.e., excluding map 
edges).  This could in principle improve the chi2.

- Lero is getting red. chi2 of ~2 for the M82 map.  This needs to be 
investigated further.

_____________________________________________________________