Bug in redshift distribution especially affecting the first source bin

LSSTDESC / augur

DESC forecasting and inference validation tool

MIT License

3 stars 2 forks source link

Bug in redshift distribution especially affecting the first source bin #51

Closed heatherprince closed 3 weeks ago

heatherprince commented 3 weeks ago

Source bin 1 has a nonzero value at z=0 which does not match the SRD redshift distribution implemented by Niko and Paul. It led to large discrepancies for power spectrum, especially for combinations with the first bin. I think this is caused by how the Gaussian filtering is implemented in Augur.

heatherprince commented 3 weeks ago

The default mode for scipy's gaussian_filter is reflect, which is going to do weird things at the edges of the array. I think we want constant with zeros.

arthurmloureiro commented 3 weeks ago

Benchmarked against @paulrogozenski notebooks in the forecasting repository.

There is a normalization factor that causes them to be a bit different but this is not relevant...

old augur: