Affiliation
American Association of Variable Star Observers (AAVSO)
Tue, 05/15/2012 - 02:14
Sometimes I take 10-12 images of a faint variable in order to stack. However, I frequently have an NSV or ASA variable in the same field that is bright. My practice is analyze each image with a bright variable and then bin the results. However, I seem only to have an option for binning 5 images. So, for example, with 12 images I turn up with 2 bins of 5 and 1 bin of 2. Is there a way to change this so that I can specify the number of binned images?
Ed
Hi Ed:
I'm a little confused by your use of the word bin rather than stack?
I just stacked 16 images, got one image and analyzed it with my normal sequence.
Can you be a little clearer about what you tried to do?
Ken
Hi Ken:
If you do a time series there is an option to bin your images. If you select "1" than all results are reported individually. But you can select 1-5 bins. So with 10 images and 5 bin you get two reports, each being single reports using 5 images. If you do so you will see your uncertainty decreases. So, I think it is good practice to bin when you have several images of a bright (SNR>100) variable you don't expect to vary over the period of observation.
Naturally if its faint I stack. And if it is variable over the series of observations (an EB for example), then I bin "1" and report all data individually. Finally, if I don't know the nature of the variable and I see some systematic variation (std of comp low, std of variable high) I report all images individually.
Cheers, Ed
Ed:
Thanks for the careful description of your process. I immediately jumped to the conclusion that all you needed to do was to stack (not time series/bin) all the images for the faint variables in the same images. I probably still would have analysed the faint targets that way BUT the efficiency of "instant" binning/refreshing is hard to argue with. ;-)
I suspect, as I'm sure you do, the reason for having only a choice of bins from one to five was a design thought. A time series analysis implies the expectation that the target will vary in a "short" time frame and thus one would not want to combine too many images in a group/bin. Of course, your desire was to add another process for the faint targets in the same image set.
Perhaps Geir "could" easily change the bin options to 1-n as opposed to 1-5. My question is how useful that would be when the stacking alternative is quite easy to do also?
SO, let's see how many users read this and give a preference??
Ken
Hi Ken:
Just back from vacation. You point on terms is on target. So, I have been examining the time series and options. From what I can see, if you wish to report several observations as a single data point (as in averaging 4-5 images) and you do not expect the target to vary over the time span of the observations the binning in VPHOT must be approached with caution. The problem is not the time series per se, but the fact that there may be more than 5 images, a point that you make. Another problem is the uncertainty reported.
So, I did some exploration: A time series of any number of images will yield the same average and standard deviation of the variable as obtained by computing the same data in Excel. No problem there. However, the uncertainty reported in the AAVSO report is 1/SNR. The fix, as long as you have 5 or fewer images is simply to select the option of "fixed" in the report option "Error Estimate" and report the actual standard deviation from the results. Of course, this works whether you expect the variable to vary over the time period of observation of not, just pick the appropriate uncertainty.
The problem only crops up when you have two or more variables in one images. For example, I have a faint one for which I need to stack 15 images to obtain an acceptable SNR and a bright one (SNR>100 in each image) for which I wish to average (having looking up its class and determined that I do not expect it to vary over the time period of observation). In this case, I cannot bin inte bright one: it would require three binned reports. So, I have been generating a report of all 15, exporting it to Excel and computing the median time, average and Std Dev of the variable. Then I delete all but one measure, paste in the average, SD and time and add the N=15 to the comments. Laborious, but I think it is "best practice."
Naturally, I am full of beans after completing the "Uncertainty" course (in fact, I am now editing all my 2012 observations). And I have been a fan of VPHOT since its Photometrica days. So I emailed Geir suggesting that he might wish to think about options for reporting uncertainties in VPHOT. As a newbie in the uncertainty business, I am not the person Geir should be talking to if changes are made.
Cheers, Ed