Summary of chop-nod analysis telecon of July 15, 2009

Giles, Lero


_____________________________________________________________

Lero's analysis of M82:

- we discussed Lero's e-mail to sharp-software sent July 10th

- Lero finds that the long (~ 18%) vectors correspond to I/sig-I < 2

- some things to try next:

divide data into 2 bins and see if we get detections of any reasonable 
significance in the southeast where we are finding post-inflation 3-sigma 
vectors.  Note: error inflation is not possible for this test.

run polsharp on the old "sharpcombine4" M82 data to investigate whether 
sharpcombine4 avoids producing the large 18% vectors.  keep in mind all 
flags should be the same as for the more recent sharpcombine5 analysis.  
Note: error inflation is not possible for this test either.

at some point we have to look at the chi2 vs. bin number tests again to 
remind ourselves of the results (we think they were very encouraging)

- some other things to keep in mind

if we are sure that we want to use I/sig-I to cut on then we have to be 
sure we believe that the sig-I are not being *underestimated*.  Looking at 
the red-chi-2 for I is a good way to do this.  (But we are not at this 
point yet.)

if we get only a small number of detections out of many points sampled we 
have to remember that knowing our significance (i.e. are they really 
3-sigma) is important, since one expects some detections by statistical 
fluke. 

____________________________________________________________