search

hippke / tls

Transit Least Squares: An optimized transit-fitting algorithm to search for periodic transits of small planets
MIT License
48 stars 24 forks source link

Uncertanties in the ephemerides, transit duration etc.? #74

Closed MilevaE closed 5 years ago

MilevaE commented 5 years ago

Hi, using TLS I found an hiden signal which: 

Period 1.32471 d at T0= 1545.5887359991175
Period 1.32471 d at T0 (JD) = 2458545.588735999
18 transit times in time series: ['1545.58874', '1546.91345', '1548.23816', '1549.56288', '1550.88759', '1552.21231', '1553.53702', '1554.86174', '1556.18645', '1557.51117', '1558.83588', '1560.16059', '1561.48531', '1562.81002', '1564.13474', '1565.45945', '1566.78417', '1568.10888']
Number of data points during each unique transit [ 9.  9.  9. 10. 10.  9.  9.  9.  0.  0.  9.  9.  0. 10.  9.  9.  9.  9.]
The number of transits with intransit data points 15
The number of transits with no intransit data points 3
Transit depth 0.99711 (at the transit bottom)
Transit duration (min) 18.79655
Transit depths (mean) [0.99509769 0.99837714 1.00028289 0.99666704 0.99655364 0.99587185
 0.99418039 0.99763466        nan        nan 1.00194123 0.9938278
        nan 0.99543035 1.00183443 0.99760156 0.99933798 0.99653153]
Transit depth uncertainties [0.00223717 0.002463   0.00280888 0.00213923 0.00265328 0.0025611
 0.00262064 0.00359985        nan        nan 0.00487221 0.00243405
        nan 0.0030145  0.00330284 0.0018631  0.0010523  0.00189251]
SNR 19.49318758226711
FAP 8.0032e-05
SDE 28.284621876567176

Which looks promising at first, isn't?. But there are few things I don't know how to manage. For example, If I want to plan observations of this potential candidate, How can I know the uncertainties of the ephemeris? It could be possible to have errors in the transit duration somehow? It seems that I never obtain the same than the ones provides by TESS for known planet candidates ..

Thanks a lot!

hippke commented 5 years ago

Hi Mileva, You can get an estimate for the period uncertainty from results.period_uncertainty. However, before you get too excited, you should thorougly vet your signal. Often, transit-like signals are caused by systematics (stellar, instrumental etc.). As an example of useful metrics, see Figure 1 in this paper. Can you make and share similar plots? I'm happy to assist further.

MilevaE commented 5 years ago

Hi Hippke!, thanks a lot for the link to the paper. I read it and I tried to follow it to improve my findings. I noticed an error in one of my steps, and now the results look less promising: 

Period 1.32471 d at T0= 1545.5887359991175
Period 1.32471 d at T0 (JD) = 2458545.588735999
Period Error 0.00100 d
Period Error 1.43514 min
18 transit times in time series: ['1545.58874', '1546.91345', '1548.23816', '1549.56288', '1550.88759', '1552.21231', '1553.53702', '1554.86174', '1556.18645', '1557.51117', '1558.83588', '1560.16059', '1561.48531', '1562.81002', '1564.13474', '1565.45945', '1566.78417', '1568.10888']
Number of data points during each unique transit [ 9.  9.  9. 10. 10.  9.  9.  9.  0.  0.  9.  9.  0. 10.  9.  9.  9.  9.]
The number of transits with intransit data points 15
The number of transits with no intransit data points 3
Transit depth 0.99711 (at the transit bottom)
Transit duration (min) 18.79655
Transit depths (mean) [0.99509769 0.99837714 1.00028289 0.99666704 0.99655364 0.99587185
 0.99418039 0.99763466        nan        nan 1.00194123 0.9938278
        nan 0.99543035 1.00183443 0.99760156 0.99933798 0.99653153]
Transit depth uncertainties [0.00223717 0.002463   0.00280888 0.00213923 0.00265328 0.0025611
 0.00262064 0.00359985        nan        nan 0.00487221 0.00243405
        nan 0.0030145  0.00330284 0.0018631  0.0010523  0.00189251]
SNR 3.517800689468416
FAP 8.0032e-05
SDE 10.80704973776568

As you can see, the SNR is too low.. :( By the way, which do you think should be the minimum SNR to consider a signal as a potential candidate? Let me show you in any case the rest of results I obtained, I would like to know your opinion if possible:

The detrended lightcurve (using WOTAN), after the removal of the two existing planets reported by TESS (by the way, I found them with better SNR and obtain the same result in period, transit duration etc. So, fantastic work done by TLS):

detrended

The periodogram:

SDE

The entire set of transits of this potential candidate:

full_transits

The phase folded:

phase_folded

The transit depths:

Transit_depths

hippke commented 5 years ago

Hi Mileva,

Congratulations! You have most likely not found a new planet, but you have a working pipeline to find new planets! That's very good.

The most commonly used SNR threshold in the literature is ~ 7, as it gives (assuming white noise) less than one false positive over the Kepler mission. Usually, SNR>10 looks visually convincing to people, which is not unimportant if you plan to publish your results. SNR~ 3 is, as you suspect, too low to be considered any further.

Don't let that discourage you. Keep searching! Michael

MilevaE commented 5 years ago

Hi Hippke! yes, sure, I will keep searching :) thanks for your assistance!