Closed SimonKrughoff closed 10 years ago
Here are some examples of what this metric can produce. A user can specify the min and max periods to test in the config. I've set it up so that the period varies smoothly with RA from minimum period at RA=0 to maximum period at RA=360. Sampling for each point is taken from OpSim assuming a standard field size of radius 1.75deg. I should note that I used ops1_1140_sqlite.db which I downloaded from https://confluence.lsstcorp.org/display/SIM/Cadence+Workshop+Simulated+Surveys.
Test 1 -- min=100days, max=300 days r-band only:
Test 2 -- min=100 days, max=300 days all bands:
Test 3 -- min=10 days, max=40 days r-band only:
Test 4 -- min=10 days, max=40 days all bands:
I think the "Slow Variables" group at the cadence workshop would like to comment on this - there was some discussion about periodograms etc there. If the code runs I say merge it, and then advertise the new code with these examples to the cadence email list for comment. Where shoudl these example plots live? This repo's wiki?
On Tue, Sep 30, 2014 at 4:14 PM, SimonKrughoff notifications@github.com wrote:
Here are some examples of what this metric can produce. A user can specify the min and max periods to test in the config. I've set it up so that the period varies smoothly with RA from minimum period at RA=0 to maximum period at RA=360. Sampling for each point is taken from OpSim assuming a standard field size of radius 1.75deg.
Test 1 -- min=100days, max=300 days r-band only: [image: 100_r] https://cloud.githubusercontent.com/assets/945715/4467978/e8f197b8-48f6-11e4-9129-50e23db56c15.png
Test 2 -- min=100 days, max=300 days all bands: [image: 100_all] https://cloud.githubusercontent.com/assets/945715/4467987/0895bebe-48f7-11e4-8b12-1a884dcb2877.png
Test 3 -- min=10 days, max=40 days r-band only: [image: 10_r] https://cloud.githubusercontent.com/assets/945715/4467997/4bdc84b4-48f7-11e4-988a-af2caeb8dc8f.png
Test 4 -- min=10 days, max=40 days all bands: [image: 10_all] https://cloud.githubusercontent.com/assets/945715/4467999/59edbf78-48f7-11e4-8984-61fed0b0aba4.png
— Reply to this email directly or view it on GitHub https://github.com/LSST-nonproject/sims_maf_contrib/pull/6#issuecomment-57397540 .
I wanted to give at least Ashish a chance to comment since he initially requested this metric. Lynne was worried about the wiki getting out of date with the code which is why I put the images with the pull request, but they could also live with the issue that goes with this branch.
I didn't mention it, but there is a description of the metric algorithm in the first commit message (the commit that implements the metric). I also tried to be fairly clear with comments in the code. Let me know if there is anything that's not clear.
Thanks, Simon!
I should be able to look at this in the next day or so.
Cheers, ashish
On Sep 30, 2014, at 5:15 PM, SimonKrughoff notifications@github.com wrote:
I didn't mention it, but there is a description of the metric algorithm in the first commit message (the commit that implements the metric). I also tried to be fairly clear with comments in the code. Let me know if there is anything that's not clear.
— Reply to this email directly or view it on GitHub.
Ashish Mahabal http://www.astro.caltech.edu/~aam
Hi, Simon. This looks good. What I understand is that because you do not yet have actual variability you have substituted RA for that and used it to look for periods within the extremes provided. Is that correct? Then, when we have real data, we will use those directly in an identical fashion. -ashish
Ashish, I implemented something very general so you can put in any model for variability you want. I don't know exactly what you mean by actual variability. We have lightcurves for several types of variable sources and realistic spatial distributions for RRLy and M-flares. I was assuming that people most frequently will want to input their own models for testing specific things.
I would suggest that the output would be more useful as (fractional?) error on the determined period: (P_LS - P_true)/P_true. That would reveal any biases more clearly than the existing output.
With that output it would then be reasonable to randomize the input periods spatially. Since the cadence is not uniform in RA, my concern is that the existing linear period ramp with RA means you don't sample all the periods in the Galactic Plane, for instance.
@ebellm I totally agree. This was not meant to be scientifically interesting at all. I was hoping that it would be a template for people to do interesting things. I can implement a version of this that outputs the error, I went with this version simply because it is more obvious that it is doing the right thing when you look at the figures. If there is a specific use case that people would find interesting, I can certainly help out with coding it up.
I'm going to merge this today. If anyone has suggestions for how to make this more generally useful, please let me know.
This is a pull request to address #5.