You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It may be that we should make sure the fitting is robust to there being NaN values in the PRF. At the end of the day, if the PRF has NaN values, we should just ignore them with the fitting and fit only where there are valid model and data values?
I don't understand why one needs code to deal with NaN values in the PSF / PRF.
That images can contain NaN values because of e.g. cosmic rays or dead pixels or whatever I get, but can't one simply ignore stars with such problems when estimating the PSF / PRF from data. This way the PSF / PRF would never contain NaN values.
Do we really want to ignore the whole stars when there is one NaN pixel in its PSF? Surethe fluxerror may reflect that it had some problems, maybe also introducing a badpixel or similar flag. Or I may just misunderstood it, and you mean to ignore the star when the fitting returned with NaN?
Oh, I'm not suggesting ignoring the whole star - I just mean that when I tried, one NaN value in one pixel could mess everything up, so I'm just suggesting being more robust to NaN values and treating it like missing data (rather than having it mess up the calculation).
Activity
cdeil commentedon May 21, 2014
I don't understand why one needs code to deal with
NaNvalues in the PSF / PRF.That images can contain
NaNvalues because of e.g. cosmic rays or dead pixels or whatever I get, but can't one simply ignore stars with such problems when estimating the PSF / PRF from data. This way the PSF / PRF would never containNaNvalues.Related: #63
bsipocz commentedon May 21, 2014
Do we really want to ignore the whole stars when there is one NaN pixel in its PSF? Surethe fluxerror may reflect that it had some problems, maybe also introducing a badpixel or similar flag. Or I may just misunderstood it, and you mean to ignore the star when the fitting returned with NaN?
cdeil commentedon May 21, 2014
I'm not sure ... let's discuss on the hangout tomorrow, OK?
astrofrog commentedon May 21, 2014
Oh, I'm not suggesting ignoring the whole star - I just mean that when I tried, one NaN value in one pixel could mess everything up, so I'm just suggesting being more robust to NaN values and treating it like missing data (rather than having it mess up the calculation).
pkrskr commentedon Mar 18, 2016
Hi!
I wish to be of some help here. Please provide me with any relevant pointers if possible.
larrybradley commentedon Nov 23, 2016
This was implemented in #159.