Affiliation
American Association of Variable Star Observers (AAVSO)
Sat, 07/14/2012 - 17:32
Hi,
I am in the process of learning VStar, but am having difficulty finding documentation pertinent to the latest version. I did view the 2010 presentation by David Benn, but there have been changes made to the program since then.
Items I need help with specificaly are period search, returning to a previous page, polynomial fit (I never did understand the meaning of this), and a few others. If there is any written documentation I can use to get started, I can then ask more specific questions about the items with which I am still having issues.
Thanks,
Keith Graham
Keith,
Unfortunately, we do not yet have one good source for documentation on using VStar for data analysis. This is in work, but I don't yet know when it will be available for use.
Meanwhile, you can take a look at some of David Benn's excellent blog posts on VStar which might be of help: http://dbenn.wordpress.com/category/astronomy-science/vstar/
In his posts, among other things, David shows how VStar can be used to plot examples from Grant Foster's book, Analyzing Light Curves: A Practical Guide
I hope that helps a little anyway. Please feel free to post more specific questions when they arise and hopefully somone will be able to answer them.
-Sara
Hi Keith
Sara mentioned my blog posts, some of which cover period search with VStar.
I have just now written another post about polynomial fit:
https://dbenn.wordpress.com/2012/07/18/polynomial-fit-to-estimate-a-mir…
Please feel free to ask further questions about any of these posts.
Can you tell me more about what you mean by "returning to a previous page"?
Apologies for the lack of a user manual. I seem to be writing user manual sections by way of blog posts at the moment.
Regards,
David
OK, I now have
Hi David,
OK, I now have some specific questions in which the answers may give me a better understanding of VSTAR.
As I viewed the VSTAR video, I noticed that it assumed a knowledge of such terms as polyniomial fit, residuals, and series bin. Perhaps there are many AAVSOers who understand these terms, but admittedly I am not one of them :). But, I am willing to learn.
1) I had no idea what a polynomial fit was until I looked it up in Wikipedia. I also checked out the link you gave me, so I now have a basic understanding of it, and this may help me to make some progress with VSTAR. I am not fully comfortable with it yet, but I will get there. I have found that when playing with the polynomial fit, if the degree is too high, the program goes into a loop which is difficult to stop. Since the icons are grayed out, the only way I found to stop it is to click on Analysis/Phase Plot. Is there anyother way to stop the loop?
2) When going into phase plot, the program pre-determines the period. However, this period is not always correct. For example, the period used for SS Cyg is the orbital period, not the outburst period. This makes for a pretty meaningless phase plot (at least for me). The only way I can see to change the period is to go back to Select a File and upload SS Cyg again. Then I click the Phase plot button and manually enter a period. Of course if the period I enter is close but not exact, I would need to reintroduce a new period. But I do not see how to enter a new period without recalling SS Cyg each time. Is there any way to enter new periods without recalling a target each time? Also, this is an example of what I meant about returning to a previous page.
3) Can you (or someone) please explain the meaning of residuals and bins as they pertain to Vstar?
May I suggest that when creating the new documentation that a basic explanation of polynomial fits, residuals, and bins be included? This would be a great help for potential users like me who took our last math courses decades ago and have forgotten most of what we learned back then:). It is now apparent to me that an uderstanding of these terms as they apply to VSTAR is imperative to using it to maximum potential.
Cheers,
Keith Graham
Hi Keith
Apologies for the delay. I've been a little busy over the last several days.
I what follows, I assume you are using VStar 2.13.
Re: creating a phase plot with a new period, use the Analysis -> Phase Plot menu item to do this. After the initial phase plot creation, the View menu Raw mode and Phase plot menu items merely switch between the light curve and phase plot. You can also return to previous phase plots via the Analysis -> Previous Phase Plots menu item. So, you don't need to load the dataset again, no.
Residuals result from taking a set of datapoints (e.g. initial observations) and subtracting a model from it. The model could be a polynomial fit or a Fourier series created from a period search (see Analysis menu DC DFT items). I talk a bit about residuals here:
http://dbenn.wordpress.com/2012/07/08/cleanest-paper-examples-using-vst…
The residuals themselves can further be analysed (models created, residuals subtracted) for additional signals. This topic, pre-whitening etc needs more space to explain. Grant Foster's Light Curve Analysis book goes into some detail.
A bin refers to the size, in days or "phase steps", of the "bucket" into which observations are lumped for the purpose of taking a mean value, one per bucket or bin. Larger bins have less uncertainty/error associated with them than smaller bins, with the tradeoff of how well the light curve or phase plot is represented. A mean curve can be viewed as another kind of model. Binned means are also related to ANOVA, the analysis of variance statistical algorithm, that indicates the likelihood of the existence of a signal (vs just randomness/noise) in a light curve. See File -> Info...
In reply to your question about a runaway polynomial fit, if this happens, you should be able to stop it via the "stop hand" button at lower right of the main window. Maximum polynomial degree is something I plan to added as a preference (see File -> Preferences...). Can you tell me a) what object and JD range you loaded observations for and b) what polynomial degree you specified?
I'll certainly cover polynomial fits, modelling, binned means etc in future documentation.
Regards,
David
Thanks for
Hi David,
Thanks for your reply. Your comments did solve some of my concerns. However I am still having a few issues. Yes, I am using V 2.213.
So we are on the same page and you can replicate my problem, let’s use a plot for ss cyg with the JD range 2455500.5 – 2456000.5.
) The Raw data gives a series of 8 max-min curves that are on the order of 40-60 days apart. When I click on Phase Plot, I get a period of 0.275130 days (this is the orbital period). Clicking phase plot presents a meaningless mess (at least to me). Setting the period to 60 days gives a series of 3 maxima, albeit they are pretty sloppy. I have yet to find, by trial and error, the plot that will give one phase. I would think that VStar would somehow check the raw data , figure the correct period, and present a phase plot using that correct period but I do not see how to accomplish this. What am I missing?
2) I tried a 9th degree polynomal fit of the Johnson V observations and got some results that made no sense to me. I then tried a 3rd degree polynomial fit of the Johnson V observations, and that resulted in the run away problem. Clicking the hand in the lower right corner did not stop the process. I had to close the program to stop it. So, how can I determine what polynomial fit to use to ultimately arrive at a meaningful result. And is there something wrong with the “hand” stopping routine or am I missing something?
Cheers,
Keith
Hi Keith
Firstly, apologies for the ridiculous length of time between replies! This last week or so has been very busy with work, family, an Arduino project, and article writing.
Now my head is back in VStar space. :)
Let me start with your second question. There's a couple of parts to my answer. One is that you have run into a bug in the current release that I have just (today, along with a similar one) fixed. This will be sorted out in the next release which I don't expect to be more than a couple of weeks away. I've been slowly working towards a new release since Easter.
Am I right in thinking that you saw this problem by: loading the SS Cyg dataset, creating a phase plot, switching back to the raw data (light curve) view, creating a polynomial fit, then seeing it embark upon a trip to la la land? In any case, you are not totally missing anything; you found a bug. Sorry about that.
The next part of my answer is to give an example of a polynomial fit for one of the SS Cyg outbursts. Given the light curve:
Create a filter (binoculars on toolbar) to select Visual observations over one of the outburst time ranges:
Then create a polynomial fit on this filtered series.
to end up with something like this.
I'm not so much justifying the use of a polynomial fit here so much as showing one reasonable application of it to this light curve. You could for example, use this to estimate the time of maximum. But if you were serious about this, you'd at least narrow the range further, e.g.
After that, you'd probably think about other possibilities like a Lowess fit (see Grant Foster's book).
Now, to the first question you asked above. As you point out, a phase plot at either the advertised (orbital) period or the average outburst period results in a mess. SS Cyg is of course a CV. With a pulsating variable such as a Mira, Cepheid, or RR Lyrae star, you'd expect to be able to get a nice phase plot at the pulsation period. For an eclipsing binary, you'd similarly like to have a phase plot that shows primary and secondary eclipses. But with a cataclysmic like SS Cyg, the outbursts are not particularly periodic, so you're not going to get a clean phase plot from those. You stand a better chance of creating a filter of datapoints between outbursts and seeing what that looks like in a phase plot at the orbital period. Even then, it's not very clean.
To address your other comment about how to find a period with VStar, in general, you need to use a DC DFT (see Analysis menu) to search for periods from which to create phase plots, models etc. I can say more about that or point you to VStar examples of this if you like.
I hope this helps.
Regards,
David