I talked with Alexia today about the reconstruction results as she had the following ideas/thoughts:
Most of the correlation power is on the order of 10Mpc/h, so I am probably binning the correlation functions too big (I am going out to 40 Mpc/h in the 3D correlation functions and 40 degrees in the 2D correlation functions which corresponds to ~1.4 in Gpc/h at z = 1 and ~200Mpc/c at z = 0.1). It might be useful to bin logarithmically as there is much more power in lower bins than in higher bins.
Alexia thinks that it is possible that the smoothing parameters of the reconstruction are causing it to look too smooth and not pick up the "double peak" feature. She mentioned using the photo-zs as the priors?
Alexia is interested in me looking into why the reconstruction goes to zero at zero and also tends to blow up at the ends. She wants me to look into this more.
I realized that it is possible that the I am doing the 2D correlation functions wrong. I need to make sure there is proper wrapping around the 0-360 degrees.
Alexia agrees that it is a good idea to re-write the code to take in both data points and randoms (for the LRG catalog and the BOSS data). She said I could test this by generating randoms with my current set of code, writing them to a file, and then reading them in with the new set of code and seeing if they produce the same correlation function.
I have a lot of running and testing to do.
Most of the correlation power is on the order of 10Mpc/h, so I am probably binning the correlation functions too big (I am going out to 40 Mpc/h in the 3D correlation functions and 40 degrees in the 2D correlation functions which corresponds to ~1.4 in Gpc/h at z = 1 and ~200Mpc/c at z = 0.1). It might be useful to bin logarithmically as there is much more power in lower bins than in higher bins.
Alexia thinks that it is possible that the smoothing parameters of the reconstruction are causing it to look too smooth and not pick up the "double peak" feature. She mentioned using the photo-zs as the priors?
Alexia is interested in me looking into why the reconstruction goes to zero at zero and also tends to blow up at the ends. She wants me to look into this more.
I realized that it is possible that the I am doing the 2D correlation functions wrong. I need to make sure there is proper wrapping around the 0-360 degrees.
Alexia agrees that it is a good idea to re-write the code to take in both data points and randoms (for the LRG catalog and the BOSS data). She said I could test this by generating randoms with my current set of code, writing them to a file, and then reading them in with the new set of code and seeing if they produce the same correlation function.
I have a lot of running and testing to do.
No comments:
Post a Comment