Fri, 04/02/2021 - 08:08
I've now worked out my transform coefficients. I was thinking that in order to get confidence in the results the way to do it would be to carry out a photometry test on a non variable catalogue star. I do the measurements and apply the transform and I should get a magnitude which is very close to the catalogue magnitude. Would that be the way to do it?
Steve
There is a feature built into the TransformApplier app that is built into VPhot that does exactly that: apply the coefficients to a known non-variable star.
If you check the Test Coefficients box and rerun the process, TA will treat the check star as the target. There is a section then at the bottom of the output box of the detailed report comparing the known, reference, magnitudes of the check stars to what is computed based on the comp star and the coefficients.
This should give you some insight into how the transform process is working for you.
Not a lot of attention has been paid to this feature. If someone with stats knowledge could advise on how to improve the report, I'd love to hear it. The questions of interest would be how confident are we that the computed estimate matches the reference. And is there an improvement in this stat when the transform is applied.
George
I'm delighted with all this interest in transforms!
At the 2019 AAVSO meeting I gave a presentation that is timely to review.
I presented a graphical demonstration of what is happening with transforms and offer a simpler, more comprehensive approach, for computing transforms that corrects errors in earlier descriptions of the process.
The power point is available here.
I'm happy to discuss further.
George
George I love that feature. I use it every time I use TA to make sure my observations look reasonable.
Barbara
Steve,
If the plots for your transforms have low values for the residuals (all points close to the fitted function) and you have a good range of B-V values, your transforms should be good.
Using photometric standard stars as targets is a good way to do a 'real' test. Image a field of standards different from your transform field, and perform ensemble photometry on each star in turn, using the others as comps.
Gerting good results for a group of stars will give you confidence. When I do that (with DSLR photometry) most V mags differ by less than 0.01 from the catalogue values. B-V results are not quite as good.
Roy
Thanks everyone, all I need now is a few clear nights.
Steve
I have worked through my data and have results for coeffecients
I have been unable to get the Transform Generator software to work as I am not experienced with python.
Is it possible to send someone my data and have them use the TG software to determine my coeffs so I can confirm my spreadsheet calcs?
bryan
I can help.
Share with me (SGEO) your M67 images on VPhot and I'll run it through TG and send you the results.
George
bryan
apprecieate it
bryan
Hi I've been working on the AB Aur campaign and submitted transformed results, but they seem a bit different than others' results even if I check for transformation by clicking on the data point in the light curve plot. In particular, my B magnitudes seem brighter than others. May be a S/N issue with my blue images. These were done with a single transform set, i.e., about 50 stars from NGC7790. Anyway, I wanted to assess my transforms, so I ran TG on a set of M67 data I took last night and compared it to the NGC7790 data I was using. The coefficients are not much different, e.g., the Tbv values are 1.260 (ngc7790) and 1.249 (M67). When I average the two sets, and run "test coefficients" on my most recent B, V and R set (Jan 11-12), I get these results:
name filter Refmag VMAGINS VMAGSTD diff error transformed
Star 2 B 0.000 -3.472 12.765 12.765 0.016 yes
Star 2 V 0.000 -5.088 11.897 11.897 0.008 yes
Star 2 R 0.000 -5.216 11.443 11.443 0.017 yes
These look pretty good, i.e., less than 0.02 for B and R, less than 0.01 for V. Esp since the check star in the AB Aur campaign is not super bright.
To anyone out there with experience with transforms, is this pretty good, ok, bad, or pretty bad? I have no idea what kind of uncertainties to expect, but these seem pretty good to me, since they are at least well under 0.1. Any views??
Thanks!
Jeff