Skip to main content

Best Option to Analyze Light Curve in VStar

19 posts / 0 new
Last post
joemmonster
joemmonster's picture
Best Option to Analyze Light Curve in VStar

Hi. I just want to know what is the best option or analysis option on VStar to analyze the light curve of my target variable star. I'm confused with these 3 options: DC DFT Standard Scan, DC DFT with Frequency Range, DC DFT with Period Range. I did once try the one with frequency range and the parameters are auto-filled with initial values (which is I don't know why they're there and how come the values are that) and when the I checked the top hits and created a model, the sum of squared residuals (SSR) are comparatively low than the other options like the standard scan, which makes the model well-suited and more precise to the obtained light curve.

I hope that someone can clear this up for me. I just started using VStar weeks ago and still trying to learn from it.

WGR
WGR's picture
Option

Hello Joe

I would start with the Standard Scan, and depending upon the results, you can then investigate further with a Frequency Range or a Period Range.

Gary

dedricksond
Light Curve Analysis

Hello Joe,
I just finished the CHOICE VStar course and highly recommend it. I believe there is another one starting soon. The book by Grant Foster, Analyzing Light Curves: A Practical Guide, is a very good source according to many.
You didn’t mention what type of variable you were interested in. A Fourier transform may not be appropriate for some types.
Peroid04 may be another option for the analysis of stars that may have multiple modes.
I consider myself a novice at these things and there are many others with AAVSO that you can get wonderful help from. Keep playing!
Duane Dedrickson

David Benn
David Benn's picture
Light Curve Analysis

Hi Joe

Further to the fine advice given by Gary and Duane, feel free to tell us more about your data/target. That may allow us to provide further advice.

For some datasets, with sufficient data, by zooming in and inspecting, a rough period can be determined as input to DCDFT. Using the Mean time between minima plugin (see Tools -> Plug-in Manager) can help with this. Doing this, followed by DCDFT with period range can be a good approach.

I personally tend to use DCDFT with period range when I can since I'm usually thinking in terms of a period range.

I'm also working on an ANOVA-based periodogram plugin that's an alternative to DCDFT. 

David

joemmonster
joemmonster's picture
Thank you for your replies.

Thank you for your replies. By the way, I'm analyzing a Cepheid variable star AD CMi which, i got the data from AAVSO database and consists of 1,117 data points. By the way, what's the basis for coming up with these auto-filled values in parameters of DC DFT w/ Freq. Range?

David Benn
David Benn's picture
AD CMi

Hi Joe

You're welcome. 

Pages 65 and 66 of the VStar user manual give some background re: where these values come from. 

Okay, so I loaded all the AD CMi data from AID. According to the VStar Info dialog (so, VSX), AD CMi is a DSCT (Delta Scuti) star rather than a Cepheid, in particular a High Amplitude Delta Scuti since its amplitude is greater than 0.2 mags.

I zoomed into the Johnson V observations centred around JD ~2454425 and used the Mean time between observations plug-in (see Tool -> Plug-in Manager) to get a rough period of 0.1282 days as shown here:

VSX says 0.12297443 days. 

Then I used DCDFT with Period Range on Johnson V, supplying the parameters:

  • low period: 0.1
  • high period: 0.2
  • resolution: 0.00001

and the best I could do was around 0.124 days. Creating a filter of the consistently lower magnitude observations and repeating the DCDFT on that gave 0.12246 days. To see what observations I've filtered, see this phase plot based upon the VSX period:

Try creating a phase plot at 0.12246 for comparison.

More analysis is required but it's getting late here.

David

joemmonster
joemmonster's picture
Filtering observations.

Thank you for the explanation but I'm having a hard time trying to understand how to filter my observation. How did you do that? I set the filtering option by magnitude but I don't know what's next. I can't select the entire lower magnitude observations.

David Benn
David Benn's picture
Filtering

Hi Joe

In this case I filtered by Observer Code.

The simplest way is to select an observation in the bottom set of observations with the mouse, then create a filter with an Observer-Code-equal-to term as shown and "Use selected observation" ticked. Then when you run DCDFT, you can select Filter rather than Johnson V as the series to analyse.

David

David Benn
David Benn's picture
Filtering

Hi Joe

In this case I filtered by Observer Code.

The simplest way is to select an observation in the bottom set of observations with the mouse, then create a filter with an Observer-Code-equal-to term as shown and "Use selected observation" ticked. Then when you run DCDFT, you can select Filter rather than Johnson V as the series to analyse.

David

joemmonster
joemmonster's picture
Thank you so much!

Thank you for being such a great help in understanding VStar. I might ask for another help some time in the future. Thank you again.

joemmonster
joemmonster's picture
Finding the Period

Hi. After a long vacation from this project, I decided to continue this despite of busy schedules. I just want to ask if it is possible to come up with a single period by combining the light curves from different observers and make them lie in the same curve? Or is it just by trial and error and intelligent guessing? I did what you've said and came up with this.

David Benn
David Benn's picture
Finding the Period

Hi Joe

What JD range are you using?

David

joemmonster
joemmonster's picture
I'm using all the AD CMi data

I'm using all the AD CMi data from AAVSO Database. So probably, from 2453810.5572 to 2457065.42077. But there are gaps from the observation.

EDIT: I mean, the other should be parallel, not lie on the same curve.

PVEA
PVEA's picture
Hi David, I tried AoV in the

Hi David,

I tried AoV in the case of AD CMi and i worked pefect. Even with no idea about the real period in two steps it reached: P=0.1229745  (see the screenshot)

Cheers,

Velimir

David Benn
David Benn's picture
Hi Joe, Velimir Nice result

Hi Joe, Velimir

Nice result Velimir!

In case Joe or others are wondering, AoV is an alternative to DCDFT for period analysis that has yet to be formally released although there is an early version available via the plug-in manager. It will be made formally available (with some doco) in the next release, later this week.

I often find it useful to start with the mean time between selection plug-in, e.g.

So, a period range of 0.1 to 0.13 with a resolution of 0.000001 and 10 bins for AoV gave me a top-hit of 0.122974 days for the period. A phase plot at 0.122975 seems better. What were your input parameters Velimir and did you run AoV against all Johnson V data or a filtered subset?

Those same parameters (0.1, 0.13, 0.000001) for DCDFT with period range gave the same resulting top-hit period as I got with AoV: 0.122974

Joe, let us know whether you can reproduce this result.

David

joemmonster
joemmonster's picture
I got the same.

I got the same result for AOV with period range and DCDFT with period range but it seems that the parameters used by Velamir is precise which corresponds to the the curves being parallel..

PVEA
PVEA's picture
AoV solution variant 2

Hi David and Joe,

All steps I followed (see the attachment):

  1. Choosing all Johnson V band data – no filtering
  2. The period range and resolution 0.115 / 0.13 / 0.0000001 (one zero more than yours)
  3. Top Hit: 0.1229744 (yours was 0.122974)
  4. Examining the LC I manually choose Epoch to be 2453810.5775

It was not possible to repeat the value 0.12275 but 0.122744 harmonizes well with the data.

Regards,

Velimir

David Benn
David Benn's picture
AoV solution

Great stuff Joe and Velimir. Thanks!

I note that the 0.122975 period is within the first 3 top hits using your parameters.

Using the Current Mode ANOVA plug-in may be useful for determining which is better. Also, creating a Fourier model via DCDFT or the Fourier model creation plug-in and looking at residuals would be interesting. 

David

joemmonster
joemmonster's picture
Fourier Model

Thanks Velimir and David. I did reproduce the result. Currently, I'm trying to create a Fourier model with this setup to make predictions. I find the residuals lesser with 4 or 5 harmonics, but not the result that I want.

 

Log in to post comments
AAVSO 49 Bay State Rd. Cambridge, MA 02138 aavso@aavso.org 617-354-0484