Re: Winter Project

["Thomas Droege" <tdroege2@earthlink.net>]

> [Original Message]
> From: Stupendous Man <richmond@stupendous.cis.rit.edu>
(snip)

>   This is a good idea.  It will undoubtedly bring to notice
> a number of "bad" frames which slipped through the current
> quality checks.

I remind all that I have already done this.  When I excluded the bad
frames, it did not change the scatter of the data as a whole.  Thus my
conclusion that the problem with the present data is the variable
sensitivity of the CCD442A as determined by the transparency of the front
side gates.  Thus we are CCD limited until we are photon statistics
limited.  This limit is of order 0.05 mag.  

Having said this, there is probably a gain to be had by excluding frames
with high scatter.  It does not help the overall statistics, but it may
help the individual statistics.  I actually think it will but there will be
a lot of work required to demonstrate this.

>   Note that one result of all this work will be a mean magnitude
> for every star, plus a measure of deviation from the mean; or maybe even 
> a mean magnitude excluding outliers, plus a measure of deviation excluding
> the outliers.  Hmmmm.  That sounds like a catalog of sorts, something
> which would be very interesting and useful in its own right.

I already compute this file.  It is labeled all.t and is one of the things
generated by my "process everything" program.  I can send this file to
anyone that wants it.  It is the one that I just quoted to someone. (Arne).
It has the following star count:

One or more hits	18,169,758
three or more 		7,233,061
5 or more 		3,997,689
10 or more		1,560,290
20 or more		498,386

This program computes sigma for each star from all the measurements.  I
have another program that computes sigma after excluding a variable number
of above and below the mean stars.   I use this one for hunting for
variables so I do not presently output a list of the stars other than as
plot files for the stars that meet all the requirements.

>   If one did this for N >> 100 stars, one could then make a plot
> with mean magnitude on the x-axis, and standard deviation from 
> the mean on the y-axis: in other words, a mag-sigma plot.  Glancing
> at such a graph would tell you a lot about the uncertainty in
> a single random measurement at any particular magnitude.
>

I have been continually doing this.  At present the N I use is 10-20.  This
because the number of available stars drops off quickly above 20 for the
present data set.  About 5 more years will get me to the >> 100.  I
regularly make such plots and this kind of plot was the basis for my
conclusion that excluding the "bad" frames does not help much.   I.e.
excluding the frames with a high percentage of outliers does not improve
the appearance of the mag-sigma plot.  OK, there may be some significance
that could be extracted by computation, but the appearance does not
improve.  That I believe is due to the characteristics of the front side
CCD.  

>   All kinds of good stuff will flow from this kind of work, even
> if one picks just a small area of the sky as a subset for convenience.
>

Yep.  The data set is getting pretty big for this work.  I am thinking of
breaking it up into "hour" segments.  Adding say an arc minute on either
side to the data to solve the problem of stars on the boundary.  Zero is
another problem, I think, as my collect_stars.pl program may need work
there.   

Tom Dreoge


> [Original Message]
> From: Stupendous Man <richmond@stupendous.cis.rit.edu>
> To: <tdroege2@earthlink.net>
> Cc: <mwrsps@rit.edu>
> Date: 10/29/2003 10:36:53 AM
> Subject: Re: Winter Project
>
>
> > The plan is to find all the outliers - say all those measurements for
all
> > stars more than 3 sigma from the mean value for a star.  Then compute
the
> > percentage of outliers in each frame.  Then flag all the readings with a
> > code as above. 
>
>   This is a good idea.  It will undoubtedly bring to notice
> a number of "bad" frames which slipped through the current
> quality checks.
>
>   Note that one result of all this work will be a mean magnitude
> for every star, plus a measure of deviation from the mean; or maybe even 
> a mean magnitude excluding outliers, plus a measure of deviation excluding
> the outliers.  Hmmmm.  That sounds like a catalog of sorts, something
> which would be very interesting and useful in its own right.
>
>   If one did this for N >> 100 stars, one could then make a plot
> with mean magnitude on the x-axis, and standard deviation from 
> the mean on the y-axis: in other words, a mag-sigma plot.  Glancing
> at such a graph would tell you a lot about the uncertainty in
> a single random measurement at any particular magnitude.
>
>   All kinds of good stuff will flow from this kind of work, even
> if one picks just a small area of the sky as a subset for convenience.
>
>                                    Michael