Deleting points in the wavelength calibration step

[cosmologist]

In the data reduction manuals when talking about the wavelength calibration step, it is mentioned that when doing the fit, if some points are really far away from the horizontal line one should simply delete them... but I don't understand what that means in terms of the reduction and fitting, how can we delete points from the fitting? Isn't that kind of "cheating"? When I read it in the manual I tried to understand it but it doesn't explain it. That way I can delete so many points to get a perfect fit but how "honest" my solution will be? Do you know what I mean?

Thanks

[favilac]

Having been recently in that position, I can tell you that we must remember that the whole data chain aquisition is not perfect. From cosmic rays to hardware issues, sometimes you get data points that are not related to the object. This type of points are usually really far from the ideal line fit, so it's easy to spot them. To confirm, you can display the spectra in DS9 and see if there is a hotspot in that place.

Another problem that may arise when fitting spectra, is the case of really close double lines. When this happens sometimes the IRAF routines are not able to assign the wavelenght in an accurate way, so its better to not include them in the fit.