search

manodeep / Corrfunc

⚡️⚡️⚡️Blazing fast correlation functions on the CPU.
https://corrfunc.readthedocs.io
MIT License
165 stars 50 forks source link

Weighted correlation function #139

Open zxzhai opened 6 years ago

zxzhai commented 6 years ago

The "npairs" output from DDrppi_mocks (and DDsmu_mocks) are unweighted, so the resulting correlation functions from Corrfunc.utils.convert_3d_counts_to_cf and Corrfunc.utils.convert_rp_pi_counts_to_wp are unweighted. Perhaps the weighted correlation can be calculated by ["npairs"]*["weightavg"], or it may be more involved.

manodeep commented 6 years ago

For a pair-product, I think supplying npairs*weightavg is fine. However, I am not sure that that holds for arbitrary weighting functions. At the very least, the existing utilities should check if the result is a weighted correlation function, issue a warning and then use the npairs*weightavg to compute the correlation function.

@zxzhai While this solves your existing problem, I would like to keep this issue open until I can figure out the correct way for arbitrary weights. Is that okay?

zxzhai commented 6 years ago

Dear Dr Sinha,

That is absolutely OK with me.

I think this solves most of the "weighted" problem in correlation function. A special case might be the collision weight in small scales because some extra conditions need to be specified in the code. I totally agree with you to keep this issue open.

Thanks, Zhongxu

On Sat, Oct 7, 2017 at 6:10 PM, Manodeep Sinha notifications@github.com wrote:

For a pair-product, I think supplying npairsweightavg is fine. However, I am not sure that that holds for arbitrary weighting functions. At the very least, the existing utilities should check if the result is a weighted correlation function, issue a warning and then use the npairsweightavg to compute the correlation function.

@zxzhai https://github.com/zxzhai While this solves your existing problem, I would like to keep this issue open until I can figure out the correct way for arbitrary weights. Is that okay?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/manodeep/Corrfunc/issues/139#issuecomment-334969066, or mute the thread https://github.com/notifications/unsubscribe-auth/AMWGGIOdg-I4BDDGVEGD4upMXUlQy-dLks5sp_a_gaJpZM4PxC0- .