Coming Data

[droege@snapmail.us]

Michael Sallman has been looking at some very old data.  I think I should
try to make some distinctions about data.

Data prior to November 2002 should be considered test data.  If it can be
used, then it is OK to use it any way you want.  But it was taken during
testing and anything could be wrong.  For much of this data we have the
raw images on CDs.  It may be difficult to find good darks for this data.

Data taken from November 2002 through December 2004 we call the
engineering run.  Changes during this time were minimal and most things
were checked carefully, but this data does include uncovering an error in
the leap year calculation. We think we have corrected everything for this
error, but this and many other things should be on a big list that is
titled "Things that we have found wrong with the data that we think we
have corrected".  For most of this data we have only the dark subtracted
and flat corrected images.

Starting with the next data we will remove the "Engineering" from the run
identification.  I expect that there will be a few months at the start
where we will be uncovering errors in my interpretation of Rob's code.  We
will watch the data closely and should be able to detect any real
problems.  Most of the things that might cause an error are now automated.
 Time, for example.  Rob's code alows a nice way to automate darks and
flats and many other things that I used to do by hand.  I expect the
processing to go much more smoothly.

I am close to starting data production from TOM1.  TOM2 and TOM3 should
follow at intervals of a few weeks.  TOM4 and TOM5 will come up over the
spring and summer. I plan to spend most of the time tracking fields.  This
should make everyone happier.  It won't make the data any better, but it
will look better.  I am working now to improve the tracking so that the
stars will not wander much from the initial position on the chip.

I remind everone that this will produce a new type of variable star.  In
the past we determined that a star drifting by a CCD defect will dip in
value.  I recally that this type of "variable" was detected by Andrew
Bennett.

No data is perfect.  We just have to advise everyone who uses the data the
various types of defects that we have uncovered.  After that, it is data
user beware.

This is somewhat different from the way data is usually presented.  Data
is worked on inside a group and nothing is said about it and none of it is
made available until the "official" release of the data.  Then a paper is
written that discloses the methods used, the data is made available, and
that is that. (After this point, if often becomes difficult to preserve
everything that went into the production of the official release.  Data
tapes deteriorate over time and computer systems and software changes. 
The result is often that it becomes impossible to check anything about the
data.)

As I like to remind everyone, with tass you are "inside" the working
group.  We have a virtual paper out that says as much as we can now say
about the engineering run data.  As the final data appears we will produce
a data paper.  Since the whole world is "inside" the tass group, there
will be no "official" data release as is usually done.  Compared to the
usual experiment, tass is in a sort of continual data release.  While I
can make some raw images available, It will be just as hard to reconstruct
problems with old data.

Tom Droege