thoughts on mucap analysis reports...

["David W Hertzog" <hertzog@uiuc.edu>]

Steve and Tom,
	Congratulations on a magnificent body of work on the analysis
reports.  I look forward to the discussions in the next days.

	Tom's report had some suggestive sections and starts of
discussions (well appreciated by me), so let me ask a simple question or
two since I brought this up earlier today before I got to Tom's last
third of his report.

	With Steve's run grouping and the relatively large impurity
correction (overall), is the impurity correction applied differentially
to the groups or is it the average correction, or is none applied at
all?  I am simply asking because (is it Fig 51 in Tom's) there is a
trend from earlier runs having higher concentrations compared to later
ones and this could give a group to group inconsistency that might be
accounted for by a differential correction.

	Next.  Tom, you mention the data set of Prod48 (different TPC
voltage).  Does the ratio of impurity events follow your "hodge-podge"
expectations from the point of view of different threshold?  In other
words, this might be used to help calibrate your roughly 69% efficiency.

	More importantly (to me), the fit that is shown for the
impurities (capture plus drift) is very good but has no shown terms or
conclusions.  You mention that you don't really like what it says, but
what does it say and why should we dismiss it?  Does this take us to a
different extreme in terms of possible conclusions?  

	Next.  I agree that the "OR" and "AND" analyses of Tom are
differing by more than we might feel comfortable with, however, I don't
know if you have done the correct calculation of the allowed difference.
You used the difference in quoted final uncertainties as in a Kawall
type band.  But, that implies that all events in one sample are in the
other plus some more.  Is that as simple as it is?  Another way is to
check just how many events are in the OR sample that are not in the AND
sample (and possibly the reverse if pileup or other conditions would
eliminate some).  Then, it is this group's effect that provides the
allowed difference.  Sometimes, we found in g-2 that what appeared to be
rather small set-subset differences (few percent) translated in to 0.5
sigma shifts or more, which are indeed allowed.  The same goes for
comparing events from Steve to Tom.  You can make an estimate of the
event overlap to establish how well you expect them to agree (I think I
recall that you guys did this already of course).

	Anyway, if this numbers are already right (let's assume they
are), then it is a very uncomfortable situation to have that much of a
change from AND to OR.  It implies that even with the OR, there remains
a time-dependent effect that you cannot lower by going to an even softer
"OR" condition.  At the moment, I have no advice, but it will be an
interesting discussion.

	See you tomorrow.
	Dave