-
Type:
Task
-
Status: Open (View Workflow)
-
Priority:
Normal
-
Resolution: Unresolved
-
Labels:None
Created on 2025-02-28 09:22:26 by Didier Vibert. % Done: 0
to derive the redshift accuracy we fit a Gaussian on the Pdf around the local maxima. In fact we fit the second derivative of the log Pdf which can then be converted to the sigma of the Gaussian.
The interval used to compute this fit starts with a fixed number of samples on the logPdf and we iterate by dividing the sample number by 2 if the estimated second derivative has a bad sign, due to the shape of the peak. After a maximum of 2 iterations, if the problem persists, the estimated delta_z is forced to be 1E-3 without any warning.
We should issue a dedicated warning in this case.