re analyzing data

[jg <jg@jgws.freeisp.co.uk>]

It might be interesting for Tom to post the data as examples.

Quite frankly there's enough variables in the July data with _large_ WS
index that show little or no variablity to the eyeball save some vague
scatter around the +/- 0.1 to 0.2 mag mark that nothing more can be said
about unless someone goes out and points a CCD and scope at them for a
few hours one night as it is, most of which can not be so examined as
there ain't folk offering to do it.

Those of these I look at rarely show any evidence of periodicity, and
cases are not infrequent where one night of suggested short term
cyclicity looks just like scatter on all the other nights, and I've no
confidence in the suggested variability being real.  Scatter on any
night can be around the 0.2 mag range.

I'm also confused that some posts suggest that magnitudes of even
constant stars may not necessarily tally on different nights at the
current state of the processing pipeline (as the system is still in
development), whereas Michael R's posting re the pipeline just now
suggests otherwise.  The "constant" "check" star I used to test the data
for a potential EA star had an internight mean mag range of about 0.1
mag for instance, over and above the same sort of level intranight.

It's hard to assess any of the data re variablilty without at least 120
obs over at least three separate nights.  The lower the amplitudinal
variation the tougher the job.

Not that there's anything wrong with all this, but if follow up
potential is exceedingly limited it seems more likely that folk are
gonna be more tempted to look at higher amplitude variation objects
because they seem more likely to be real.

There are a LOT of low amplitude little quirky bits of data objects,
whose gonna test 'em all?

NOTE: this is within the context of previously unknown variability. 
It's a doddle if the object is already known variable, or suspected
variable, or even a doddle if a hot Jovian sized body causes eclipse
dips comparable in size to the error bars if there's at least a most way
accurate ephemeris to predict eclipses from.  Hindsight is magic, first
proof is a bugger.

Summary.  I'd like some specific example data of a couple of these low
WS index, low amplitude, short term periodicity, potentially variable
objects, please.


RE outliers: isn't there some sort of simple statistical test using the
distance between the median (or calculated mode?) and mean of a data run
to check for skewing outliers?  I suppose it depends how out the lie is.

Cheers

John G.

Other data analyzing opinions may vary