Skip to content

Latest commit

 

History

History
180 lines (146 loc) · 6.18 KB

expected_results.rst

File metadata and controls

180 lines (146 loc) · 6.18 KB
Raw

Testing our YSOVAR codes

We have generated artifical csv files that have the same format (at at least a very similar one) to the csv files extracted from the YSOVAR2 database. The purpose of this excercise is to

  • test that our code works correctly
  • compare our python code to Maria's IDL code

Just like the real data, one singe csv file contains many different sources and each source is meant to verify one or more of the statistics we use. Most of them are analytically generated, so we know what the result should be exactly, some are randomly generated. They might change between different run (I'll implement a constant seed of the random number generator later).

Parameters that are uninteresting for the test in question (e.g. RA, DEC or the magnitudes when we only care for the number of datapoints) are randomly generated and will also change each time we rerun the program. Also, I did not bother to set the number of digits for each item, so that e.g. ra and de might have less digits and mag1 might have more than in the csv files. Let me know if that is a problem, I can fix that.

All sources in the csv file have negative ysovarid entries to set them apart form real data. Below, we list for each source which statistics we want to verify there.

Note

All the examples below assume that no systematic error (i.e. no error-floor is added). If you do add an error floor, then obviously all values that depend on the error (chi^2, stetson) will not come out as analytically calculated.

One-color statistics

We generated several very short and simple lightcurves for different sources, that have entries in one band only.

Which lightcurves make the cut?

We ignore lightcurves below a certain number of datapoints. To check that this rejection works, here are a few (constant) lightcurves with few datapoints.

Source number number of datapoints
-998 4
-999 5
-1000 6

All other lightcurves in the simulated dataset have 6 or more datapoints (most have 50 or 100 datapoints).

constant lightcurves

These are two lightcurves that have identical entires, one of them tests the IRAC1 channel, the other one the IRAC2 channel. The followinf statistics should be compared:

  • number of entries = 6 in the band chosen
  • max = 12
  • min = 12
  • median = 12
  • weighted mean = 12
  • any quantile = 12
  • standard deviation = 0
  • median absolute devitation = 0
  • chi^2 to mean = 0
Source number band lightcurve
-1000 IRAC1 constant
-1001 IRAC2 constant

simple lightcurves

These two lightcurves have identical entries again. The following statistics should be compared:

  • number of entries = 6 in the band chosen for -1002 and -1003, 100 for -1004
  • max = 13
  • min = 11
  • median = 12
  • weighted mean = 12
  • standard deviation = 1./sqrt(2) = 0.7071 (taking dof = 1) (not for -1004)
  • median absolute devitation = 0.5 (not for -1004)
  • reduced chi^2 to mean = 50.0 (only approximate for -1004)
  • 10%-90% spread of the distribution = 1.6 (only for -1004 - depending on how your algorithm treats the values on the boundaries, this might vary +- 0.02)
Source number band lightcurve
-1002 IRAC1 linear brightening
-1003 IRAC2 linear brightening
-1004 IRAC1 linear brightening

lightcurve that would show if mag and error are reordered

This lightcurve would show if errors are messed up with respect to the times. Thus, most important is a comparison for quantities that depend on both, magnitude and error:

  • reduced chi^2 to mean = 0.9665139
  • weighted mean = 12.31796
Source number band lightcurve
-1010 IRAC1 jumps up and down

Periodicity

All lightcurves here are generated with a sin-wave, most have some added noise, so the expected results can vary a little. We add the noise, because some implementations of Lomb-Scargle fail, if the solution is exact (the peak of the periodogram could be infinite).

Source number band Noise? expected results
-1500 IRAC1 no period = 3
-1501 IRAC1 yes period = 50, peak is significant
-1502 IRAC2 yes no significant peak
-1503 both yes period = 3 in IRAC1, =5 in IRAC2 IRAC1 is more significant

Merging lightcurves with data in different bands

The following lightcurves have different observation times in IRAC1 and IRAC2 (e.g. IRAC1 before IRAC2, IRAC1 after IRAC2). The table gives the expected number of datapoints where photometry for IRAC1 and IRAC2 exisits within 12 minutes, 0.01 days, or 0.05 days.

Source number 12 min 0.01 days 0.05 days
-2000 none none none
-2001 50 50 50
-2002 8 10 50

Two-color statistics

Here are a few sources to check shapes in the color-magnitude-diagram.

Source number color change CMD shape
-2500 None flat
-2501 random point cloud
-2502 reddening line

And now the Stetson index

Source number Stetson index
-2700 0.00
-2701 0.00
-2702 10.1012
-2703 10.1012
-2704 -10.1012