meridian flip issue - not lining up the data nicely after detrend

[gboyle]

mmm. i'm not sure if these images are uploading...but in case they are..
i had expected that my post meridian flip data - except the raw series for the target - but the  detrended ones, would line up if i detrended for the flip.  as i hope can be seen in the images they do not.

Am I missing something?

best
gavin

[gboyle]

this was the other image i meant to add

[gboyle]

[karenacollins]

Hi Gavin,

I'm not really sure what may be going on, but I noticed the transit fit is not converging (the red boxes around the fit results). That may be because the inclination is "locked". Try unlocking it and see if that helps. There are also a lot of detrending parameters selected. In general, I'd only use meridian flip with one other, or at most two others. The time series is really not long enough to support more than that (especially with the limited OOT.

As far as the comp light curves, the jump in the data is fairly large and may just be outside the range AIJ is checking. However, there are two detrend modes, so go to Multi-plot_Main->Preferences and try the opposite mode that you have set to see if that helps:



Karen

[gboyle]

Hi Karen

thanks for your reply.
your detrend mode adjustment took care of much of the "jump" in the pre and post flip data - the comps still seem to be a little off the trend line but not the problem it was.

thanks for the steer on the over parameterised model - i need to find a way to better understand what is going on in the modelling so that i would intuitively recognise such errors.

i had locked inclination as i know the parameter from the discovery paper but have unlocked it .

attached then is the adjusted plot.

unfortunately the data for the target star (had to show scaled to 50% just to get it all on page) seems more randomly variable than the comps and so i am not getting a typical curve - we should have expected a transit of 8-9mmag...

my SNRs appear very high ( around 1000) and the adu count for the target actually gets into non linearity/saturation level.  when i look at images about number 7/8/9 of the run the rel_flux for the target drops when it does not for the comps - in these images the target is saturated which i think would cause that (ie if the sky can be a little brighter for a minute but the ccd cannot read the star any brighter the rel value of the star falls)

lesson here is that I should drop the exposure level - by 50% in this case and maybe even more - this would also give me many more points on which to base my models...

best
gavin

[karenacollins]

Yeah, saturation causes all kinds of trouble. I'll bet you have a much easier time with the fitting and detrending with a new dataset.