Add Treecor cross correlation for clusters vs sources

[eduardojsbarroso]

No description provided.

[eduardojsbarroso]

@combet I created this stage and it seems to be working but I have a couple questions. I checked that in the notebook example you provided here, you create already source galaxies behind the cluster. Since now we have source_tomo_bins and cluster_tomo_bins that might not be the same, I made a filter to see if the bins overlap. If there is any overlap, I do not compute the cross correlation.

But then I took the cross correlation of all source bins for a given cluster bin and computed the mean profile, IDK if this is the right approach (here is a pseudocode to explain what I am saying)

For cluster_bin in cluster_bins:
    For s_bin in source bins:
        tg_signal += s_bin.xi * npairs
    np.mean(tg_signal)

Also by doing this, I do not know how to properly compute the covariance. But I have a first code implementation that seems right, besides the cov computation. @marina-ricci If you have any thoughts on this, let me know.

An example of what I did is in TXPipe/notebooks/cluster_counts/gt_treecor_test/Run_CL_pipeline_treecor.ipynb

[eduardojsbarroso]

@combet I created this stage and it seems to be working but I have a couple questions. I checked that in the notebook example you provided here, you create already source galaxies behind the cluster. Since now we have source_tomo_bins and cluster_tomo_bins that might not be the same, I made a filter to see if the bins overlap. If there is any overlap, I do not compute the cross correlation.

But then I took the cross correlation of all source bins for a given cluster bin and computed the mean profile, IDK if this is the right approach (here is a pseudocode to explain what I am saying)

For cluster_bin in cluster_bins:
    For s_bin in source bins:
        tg_signal += s_bin.xi * npairs
    np.mean(tg_signal)

Also by doing this, I do not know how to properly compute the covariance. But I have a first code implementation that seems right, besides the cov computation. @marina-ricci If you have any thoughts on this, let me know.

An example of what I did is in TXPipe/notebooks/cluster_counts/gt_treecor_test/Run_CL_pipeline_treecor.ipynb

I gave up on using the rlens metric since I do not have sources with this metric as well and the cross correlation does not work if the catalogs are not in the same metric. I also do not know how to generate the center_patches, so this stage is an initial implementation

[combet]

@combet I created this stage and it seems to be working but I have a couple questions. I checked that in the notebook example you provided here, you create already source galaxies behind the cluster. Since now we have source_tomo_bins and cluster_tomo_bins that might not be the same, I made a filter to see if the bins overlap. If there is any overlap, I do not compute the cross correlation.
But then I took the cross correlation of all source bins for a given cluster bin and computed the mean profile, IDK if this is the right approach (here is a pseudocode to explain what I am saying)

For cluster_bin in cluster_bins:
    For s_bin in source bins:
        tg_signal += s_bin.xi * npairs
    np.mean(tg_signal)

Thanks @eduardojsbarroso. I don't know either if this is the right approach (I'm not very familiar with this) or if there are some extra subtleties to consider (pinging @tae-h-shin who might have some insight).

Do you have a way to merge the 3 catalogues from the 3 source tomographic bins into a single one, recompute, and compare the results to the averaging you do above? This would give us a hint if this is equivalent.

[eduardojsbarroso]

@combet I created this stage and it seems to be working but I have a couple questions. I checked that in the notebook example you provided here, you create already source galaxies behind the cluster. Since now we have source_tomo_bins and cluster_tomo_bins that might not be the same, I made a filter to see if the bins overlap. If there is any overlap, I do not compute the cross correlation.
But then I took the cross correlation of all source bins for a given cluster bin and computed the mean profile, IDK if this is the right approach (here is a pseudocode to explain what I am saying)

For cluster_bin in cluster_bins:
    For s_bin in source bins:
        tg_signal += s_bin.xi * npairs
    np.mean(tg_signal)

Thanks @eduardojsbarroso. I don't know either if this is the right approach (I'm not very familiar with this) or if there are some extra subtleties to consider (pinging @tae-h-shin who might have some insight).

Do you have a way to merge the 3 catalogues from the 3 source tomographic bins into a single one, recompute, and compare the results to the averaging you do above? This would give us a hint if this is equivalent.

Thank you for the feedback. Yes I can do this, I was just thinking that I do not know if it is the right approach also because we would have a calibration factor for the shear of each bin, so if I do a cross correlation with "1 bin", it is also a good way to do it. I can try this and also compare to the tangential shear from the individual profiles to see if all is making sense, but I will have to this in a couple of weeks since now I have some other urgent matters to take care :/