incorrect period guess from exofop

[avivajpeyi]

EG TOI 1982

raw data	init fit	phase

• from the left most plot, it looks like there is only one transit bw 2323-2325 days
• could there be another exoplanet in the data?

[avivajpeyi]

This looks like a system with data for only a single transit

#29

[avivajpeyi]

Add 2180 as a TOI to test

[avivajpeyi]

The third looks like an initialization problem. I'd say that the initial guess for the period is wrong. Can you double check the non-folded version to make sure that it makes sense?

Yeah dan was right about the initial period being off -- I think we are estimating a shorter period than it actually is.

See if the same error occurs for 2180

[avivajpeyi]

TOI	Period (days)	Epoch (TBJD)	Depth (ppt)	Duration (days)	Single Transit
1982.01	8.58635	1602.58	5.26	0.222833	False
2180.01	0	1830.79	2.4	1.30892	True

Regarding TOI 1982

• period guess of 8 days is wrong (cant see transits in 8 day periods)
• incorrectly labelled as "not a single transit system" our single-transit check uses the period (incorrect here) and data available

Regarding TOI 2180

• correctly labelled as a single-transit system
• do not have a period guess: causes issue with our single-transit section of the PE model (fails the testpoint check)
  (made note of this in Single Transits #29 )

[avivajpeyi]

@dfm -- what should we do in the case where SPOC's initial fits may be off?

Data for TOI 1982:
https://exofop.ipac.caltech.edu/tess/target.php?id=437329044

Because the period ~8 we have marked this as a system with multiple transits in the data, and accordingly set our priors:

[avivajpeyi]

Another TOI (TOI 2148) with an incorrect initial guess:

ExoFOP page suggests a period of 23 days

This SPOC fit may have been made during the 1st time this TOI was observed! See old plot when there was just data for ~25 days. Here a period of around 23 is resonable

[avivajpeyi]

Some follow-up suggestions on how to deal with this documented here: https://github.com/dfm/tess-atlas/wiki/Nov-24,-'21

What do we do when we have the wrong initial guess?

• Test: quality of fit for odd and even eclipses
• Even number and odd number eclipses
• At MAP value -- are later eclipses much worse fit?
• If 20sig outliers -- something probably off
• We can do this with the MAP value (before PE)

[avivajpeyi]

• find more clear examples of this issue

[avivajpeyi]

Another example: TOI 1134

• SPOC page
• No initial period provided (updated Oct '21)--> thus we assume it is a single-transit system
• looking at the data, there are 4 transits

• Even though our analysis' priors + initial guess are way off, the posterior productive plot looks reasonable:

• Our GP 'masks' out the other 3 transits (below we zoom into the 2nd half of the data after the gap)

[avivajpeyi]

Looks like this might still be an issue!
http://catalog.tess-atlas.cloud.edu.au/content/toi_notebooks/toi_1982.html