[Aavso-photometry] stats question

[Radu Corlan]

On Fri, 3 Dec 2004, Chuck Pullen wrote:

> Michael - as I first started using MIRA by Mike Newberry, I've followed his 
> suggestions of using median combine for flats and darks.  I use the mode 
> method for background subtraction.  If there's a better way, I'd be willing 
> to change!

The median is simple and robust, but not very efficient in improving the 
snr (it has about 65% of the efficiency of the average). If you combine a 
lot of frames, median is good - otherwise the clipped mean is probably 
better.

A nice way to test is to try and combine some bias frames 
with diffeent methods and check the standard deviation of the pixel values 
across the frame - the lower it is, the more efficient is the combining 
method.

To check for the robustness (how well the combining method handles deviant 
pixels -- cosmic ray hits or star images in sky flats) take a number of 
normal frames of different fields and combine them together. A good 
algorithm should delete all the stars - so you end up with just the sky 
background. I've found kappa-sigma clipping achieved this quite easily.

As for local sky estimation, the mode is sometimes ill-defined (when the 
histogram of the sky values is "jagged" at the top). Some people preffer 
using a "synthetic mean", which is ((k+1) * median - k * mean) -- where k 
is usually 2. This works best if deviant pixels (usually stars in the sky 
annulus) are excluded from the median and mean calculations.

I have a description of frame combining methods at:

http://astro.corlan.net/gcx/html/node7.html#SECTION00750000000000000000

and of the synthetic mean at:

http://astro.corlan.net/gcx/html/node8.html#SECTION00820000000000000000

Radu

> 
> Chuck Pullen
> 
> At 10:46 12/3/04 -0600, Michael Koppelman wrote:
> 
> >What do you think is the best way to combine images like darks and flats:
> >
> >1. mean
> >2. min/max clipped mean
> >3. sigma clipped mean
> >4. median
> >5. other
> >
> >Do you think the same "best" method is best for bias, darks and flats or 
> >is one method better for flats and another method better for darks?
> >
> >I use min/max clipped mean in most cases.
> >
> >Cheers,
> >Michael Koppelman
> >http://www.lolife.com/astronomy/
> >
> >_______________________________________________
> >Aavso-photometry mailing list
> >Aavso-photometry at mira.aavso.org
> >http://www.aavso.org/mailman/listinfo/aavso-photometry
> >
> >
> 
> _______________________________________________
> Aavso-photometry mailing list
> Aavso-photometry at mira.aavso.org
> http://www.aavso.org/mailman/listinfo/aavso-photometry
> 

-- 

-------------
Radu Corlan       Snail Mail: Bucuresti sect. 1, 
rcorlan at pcnet.ro  str. Argentina nr. 28, Romania

   You can still escape the "Gates" of Hell!   
                 Use Linux!