Re: Error was GSC 748-1618 period

[Tom Droege <tdroege2@earthlink.net>]

Michael K,

Sorry to beat on this point, but we are comparing two different sets of 
data.  When this is done, it is really important to understand the errors 
in order to see if there is any significant difference.

I think what you are quoting below as "average error in V on each night" is 
the photon statistics.  This is just one of the errors.  A second error is 
the night to night consistency.  As I noted in a previous message this is 
of order 0.05 mag at this magnitude.  A third is the zero point offset.  I 
have no idea what this is for this data.

For the Koppelman data we have the same errors.  I will believe something 
like 0.01 mag for the point to point consistency on the same night.  It 
might even be that good for night to night, but this would take an 
experiment and some data reduction to make me a believer.  Then there is 
the zero point offset.  I would like to see something to support that it is 
also within 0.01 mag.  From my reading this is a very difficult level to 
achieve.  What say Arne?

So we start with the tass errors at 0.05 mag and add some unknown quantity 
in quadrature to get the error bars on the tass data.  We then put some 
realistic value on the Koppelman errors.

When we do both, then my bet is that the error bars will overlap.  This 
will then say that there is no significant difference in the two curves.

What may be left is the difference in shape.  This may be worth 
investigating.  It may well be that there is a big difference in time 
between the tass data and the Koppelman data.  The data does not (yet) rule 
out a shape change caused by the star undergoing some sort of phase change 
in the interval.  So the shape could be interesting.

It would be nice to look at the curve at:

http://www.lolife.com/astronomy/gsc748-1618/7481618_phase_3.jpg

and be able to say that there is a difference in the two data sets.  In 
practice, one really has to be able to understand the errors involved.  I 
know that I don't understand my errors well enough to say there is a 
difference.

On the positive side, this star is getting a pretty thorough workout.  I 
bet it has already received more thought than the average "classified" 
star.  That is something that can be done in a group like tass.  Experts 
(and experts in training) can gather here and argue over the 
characteristics of a star as determined by some data.  This can encourage 
further measurements that test possible theories.  At time goes on, I 
expect that we will get very good at this.

We are also shooting at a moving target.  As time goes on, classifications 
change.  The more data and stars we have of each type, the better we will 
be able to classify.  There is no final answer.

Tom Droege

At 01:17 PM 2/17/03 -0600, you wrote:
>According to the TASS data, the average error in V on each night was:
>
>2452286 0.023
>2452311 0.040
>2452312 0.025
>2452313 0.029
>
>311 is the night that seems like it might be cloudy or something.
>
>I got some more data last night. I created a new plot where I binned my 
>data, added 0.01 error bars to my data and 0.03 error bars to the TASS 
>data. The new plot is here:
>
>http://www.lolife.com/astronomy/gsc748-1618/7481618_phase_3.jpg
>
>If I average the 3 periods that folks have been kind enough to send me, we 
>are still sitting at about 0.77737 days.
>
>Yes, Arne, all my data is V band.
>
>TASS V-I for the variable star:
>
>Average 0.76
>Min     0.65
>Max     0.94
>Std Dev 0.06
>
>Michael Koppelman