It's just statistics

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

Sebastian and all,

It  is just statistics.  In a large data set there will sometimes be large
variations of even a constant star.  The +/- 0.2 spread overstates the tass
error, I believe by a factor of about two.  Yes, you can find examplse
where the scatter is more.  You can also find examples where the scatter is
less.  The errors I quote S&T10 are for the whole data set.  This is useful
to place error bars on any measurement.  Error bars never mean that every
error is within one sigma.  

Note that I could easily take data with very low scatter.  This can be done
by tracking the sky for hours.  The result can go below 0.01 in spread.  
We have lots of data taken like this.  Such data is misleading, I think. 
It looks great!  But it hides the photometric error.  I have chosen to take
present data in a way that does not hide the error.  The error seen is
pretty much the error in the absolute photometry.  OK before you all jump
on me, I have not yet demonstrated this.  But I have done enough work that
I think the error curve of S&T 10 will pretty much be the error of the
absolute photometry when we get done with it.  I think not much different
than what it is today.

OK, for detecting minute to minute variations, the tass data is not so
good.  We only take one measurement a night.  ;^) But if you want to know
how bright the star is in V or I and know what the limit of the error is on
any one day, then I think the tass data might be very useful.  

I plan to continue covering the whole sky with a measurement per night as I
have been doing for several years.  When this is completed, and when I get
a few more telescopes built, I will then turn them loose tracking the sky
for 3-4 hours a night.  This will generat great looking data from the
standpoint of scatter.  This data can then be tied to the earlier data with
some confidence that that we know the zero point.  From time to time we
will continue to take all sky data with several telescopes to pick up the
really long period variables.  

Tom Droege




> [Original Message]
> From: Sebastian Otero <varsao@fullzero.com.ar>
> To: <aavso-discussion@mira.aavso.org>; <tdroege2@earthlink.net>
> Cc: <tass@listserv.wwa.com>
> Date: 8/30/2004 4:28:12 PM
> Subject: Re: [AAVSO-DIS] IO Aurigae and the STARE eclipser Aur0 3503
>
> > > TASS data have a +/-0.2 mag. scatter so there is no way to tell if
there
> > are
> > > eclipses or not since the amplitude discovered by STARE is only 0.07
> mag.
> >
> > Could you explain how you arrived at the scatter above since it does not
> > appear to be consistent with the data in our S&T note 10 at
> > http://www.tass-survey.org/tass/showtell/st0010.html
>
> Tom,
>
> Brian has already answered the question.
> This happens all the time in different fields (not crowded).
> TASS observations for some stars are good and for others are useless. I
> don't know how this comes to be and why it happens. In the example above
it
> becomes pretty evident what makes the errors to stand out easily: almost
> simultaneous measurements giving very different results. It's not a cirrus
> thing or the like because it happens all the time and the pattern is : one
> observation bright , the following 0.2 or even 0.3 mag. fainter and so on.
> The big problem with this is that when there is a single observation per
> night... how can we believe in its value? It becomes more a  matter of
faith
> than anything else.... Something must be happening and maybe it can be
> corrected.
>
> I attach an even worse example I found during my eclipsing binaries
studies.
> It is a small gif so I hope noone bothers with the attachment.
>
> On the other hand, there are other stars (at the same magnitude) where the
> data can be used. So the problem is even more difficult to understand.
>
> Cheers,
> Sebastian.
>
>