Implementation of corrections for projection effects [version:1.0.11] by eduardojsbarroso · Pull Request #88 · LSSTDESC/crow

eduardojsbarroso · 2026-06-19T11:37:56Z

This PR adds support for projection effects in CROW, including a Costanzi-style projected richness model and an optional projection-induced lensing profile correction for stacked DeltaSigma predictions.

Main Changes:

Added a new crow.cluster_modules.projection_effects module.
Implemented CostanziBaseModel, CostanziBinned, and CostanziUnBinned for projected observed-richness probabilities.
Added CostanziLensingBias / bsel for multiplicative selection-bias corrections to stacked DeltaSigma profiles.
Extended binned recipes to optionally apply lensing profile corrections after stack integration.
Added projection-effect grid construction and caching to GridBinnedClusterRecipe.
Added a projection tutorial notebook: notebooks/run_projection.ipynb.
Added and expanded tests for projection probabilities, projected counts, grid caching, and DeltaSigma correction behavior.

Implementation Notes:

The richness projection model follows the Costanzi projection-effect formalism.
The lensing profile correction is explicitly restricted to DeltaSigma predictions and raises an error for reduced shear.
Radius conversion for the lensing correction is explicit: either comoving Mpc/h input or physical Mpc converted using reference redshift and h.

Add Costanzi19 projection model and binned_grid skeleton

…py. Add function comments. Clean the code by isort and black

Implement Costanzi binned and unbinned projection-effect probability calculations for Gaussian richness-mass relations. Add recipe support for projection-effect grid construction and cache handling. Update tests for binned probabilities, unbinned PDFs, mass-redshift meshing, and projection recipe grid construction.

…profiles

codecov · 2026-06-19T11:44:12Z

Codecov Report

❌ Patch coverage is 96.53179% with 18 lines in your changes missing coverage. Please review.
✅ Project coverage is 97.99%. Comparing base (82846b4) to head (2a3bdc0).

Files with missing lines	Patch %	Lines
...cluster_modules/projection_effects/lensing_bias.py	83.60%	10 Missing ⚠️
crow/recipes/binned_grid.py	87.09%	4 Missing ⚠️
crow/recipes/binned_parent.py	88.23%	2 Missing ⚠️
...row/cluster_modules/projection_effects/costanzi.py	99.31%	1 Missing ⚠️
tests/test_recipe_binned_counts.py	98.59%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #88      +/-   ##
==========================================
- Coverage   98.41%   97.99%   -0.42%     
==========================================
  Files          28       32       +4     
  Lines        1826     2341     +515     
  Branches       60       89      +29     
==========================================
+ Hits         1797     2294     +497     
- Misses         29       47      +18

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

… models This update responds to review feedback by removing the combined projection+mass-richness classes (CostanziBinned and CostanziUnBinned) from the PR. The projection model is now kept as a standalone richness-bias kernel, while the full integration over rich_tru, mass, redshift, purity, and observed-richness bins is left for a future recipe-level PR. Main changes relative to LSSTDESC/crow main: - Add crow.cluster_modules.projection_effects. - Add richness_bias.py with CostanziRichnessBias: - P(rich_obs | rich_tru) - observed-richness-bin integral at fixed rich_tru (this is useful for calculating the number counts over a observed richness bin?) - default Costanzi projection parameters (this can be removed in the future) - Add lensing_bias.py with CostanziLensingBias / bsel: - multiplicative post-stacking DeltaSigma correction - explicit radius handling: comoving Mpc/h or physical Mpc converted using z_ref and h - Wire optional lensing_profile_correction into binned recipes: - applied after stacked lensing integration - restricted to DeltaSigma - raises for reduced shear - Add projection tutorial notebook. - Add tests for: - Costanzi richness-bias probability shape, positivity, finiteness - observed-richness-bin integration at fixed rich_tru - parameter validation - lensing-bias radius conversion and shape validation - exact/grid recipe DeltaSigma correction behavior - reduced-shear rejection - Remove the earlier CostanziBinned / CostanziUnBinned classes. - Remove the temporary grid projection-cache integration from this PR.

eduardojsbarroso · 2026-06-23T07:13:22Z

+arrayLike = int | float | npt.ArrayLike
+
+
+def _as_1d_array(name: str, value: arrayLike) -> npt.NDArray[np.floating]:


I think we don't need this function. All that it is doing is calling np.atleast_1d with dtype=float.

It would be easier if all the functions just take numpy array as inputs, like the other codes.
Check: https://github.com/LSSTDESC/crow/blob/main/crow/cluster_modules/mass_proxy/murata.py

So ideally, change every function here that takes arrayLike to take npt.NDArray[np.float64]

eduardojsbarroso · 2026-06-23T07:14:07Z

+    return array
+
+
+def _selection_parameter_arrays(


If you changed all the inputs to numpy array, we dont need this function anymore as well. It is just calling the as_1d_array and checking if they have the same length, which we can do later

eduardojsbarroso · 2026-06-23T07:18:04Z

+    return parameters
+
+
+def bsel(


Is there an specific reason for this function to be outside the class?

Have moved this inside the module.

eduardojsbarroso · 2026-06-23T07:18:25Z

+    if np.any(Rcmv <= 0.0):
+        raise ValueError("Rcmv must contain only positive radii.")
+
+    A_sel, alpha_sel, beta_sel, gamma_sel, R0_sel = _selection_parameter_arrays(


Here can just be now a test to see if the parameters have the same length

eduardojsbarroso · 2026-06-23T07:19:54Z

+
+
+def bsel(
+    Rcmv: arrayLike,


If this function is in the class, and you will check my comment about parameters, it would only take Rcmv as input to run

eduardojsbarroso · 2026-06-23T07:51:36Z

+        ) * 2.0  # (n_obs, n_tru,)
+        term1 = (
+            (1.0 - fmsk) * fprj * tau + fmsk * fprj / rich_tru
+        ) * CostanziRichnessBias._exp_times_erfc(log_exptau, erfc_arg1)


self._exp_times_erfc

I think we can still use staticmethod for this function? Because we make the probability and its integral (over rich_obs) instance methods already?

eduardojsbarroso · 2026-06-23T07:51:47Z

+            fmsk
+            * fprj
+            / rich_tru
+            * CostanziRichnessBias._exp_times_erfc(


self._exp_times_erfc

Same as above.

eduardojsbarroso · 2026-06-23T07:52:07Z

+
+    @staticmethod
+    def prob_richobs_at_richtru(
+        rich_obs: arrayLike,


Make the input ndarray

No wrappers for arrayLike now.

eduardojsbarroso · 2026-06-23T07:52:23Z

+
+    @staticmethod
+    def Sprob_at_richtru(
+        rich_obs_eds: arrayLike,


NDArray and not static method anymore

eduardojsbarroso · 2026-06-23T07:53:36Z

+        rich_obs_digit = np.digitize(rich_obs_bins, bins=rich_obs_eds)
+
+        # calc
+        prob_obs = CostanziRichnessBias.prob_richobs_at_richtru(


self.prob_richobs_at_richtru

…tion directly to bsel. Use default Parameters to store and unpack parametrs. Corresponding tests are changed to reflect the changes. Only modify for the lensing bias module for now.

…as instance methods. Users are required to provide projection effect parameters when creating an object. Otherwise, it will use the default costanzi parameters (which can be changed later). Use PArameters to store and unpack parameters. Tests are changed to reflect the modifications. Use the variable name fractional_sig_pure instead of sig_pure.

- CostanziRichnessBias keeps its calibration parameters inside the object via Parameters. - Array conversion wrappers and the arrayLike alias are removed. Use direct npt.ArrayLike. - Lensing bias uses explicit radial_units with clear behavior for Mpc/h or Mpc - Restore all recipes. - Tests were cleaned.

- Removed the combined binned Costanzi implementation and kept the projection model as P(lambda_obs | lambda_true). - Updated projection-effect inputs to use npt.NDArray[np.float64]. - Added RICHNESS_BIAS_DEFAULT_PARAMETERS and LENSING_BIAS_DEFAULT_PARAMETERS. - Updated both classes to initialize self.parameters from defaults, with optional constructor overrides. - Moved lensing-bias evaluation into CostanziLensingBias.bsel(). - Replaced the radius boolean with explicit radial_units. - Updated projection tests; tests/test_proj.py passes.

yuriuno and others added 13 commits May 22, 2026 19:15

Add Costanzi19 projection model and binned_grid skeleton

8172399

Merge pull request #2 from yuriuno/prj

21eecad

Add Costanzi19 projection model and binned_grid skeleton

use statismethod in prob and Sprob. Change the file name to costanzi.…

f3813c4

…py. Add function comments. Clean the code by isort and black

delete costanzi19

5789d37

Refactor Costanzi model and add unit tests

ec8050a

Update projection tutorial notebook to use CostanziBaseModel

70cc95e

create a new dir. copy costanzi.py here. slightly modify the code.

67ae847

delete duplicated costanzi.py

7e776c9

remove redundant imports

2b64e99

implement the projection effect correction for stacked WL DeltaSigma …

768026d

…profiles

Merge branch 'main' into prj

27334d9

Run black and Isort

2a3bdc0

eduardojsbarroso assigned inonchiu and yuriuno Jun 19, 2026

github-actions Bot changed the title ~~Implementation of corrections for projection effects~~ Implementation of corrections for projection effects [version:1.0.11] Jun 19, 2026

eduardojsbarroso commented Jun 23, 2026

View reviewed changes

inonchiu and others added 5 commits June 23, 2026 17:04

clean redundant wrapper for atleast_1d and arrayLike. Rename distribu…

d6860e0

…tion directly to bsel. Use default Parameters to store and unpack parametrs. Corresponding tests are changed to reflect the changes. Only modify for the lensing bias module for now.

Update run_projection notebook with new analysis

5586b04

		arrayLike = int \| float \| npt.ArrayLike


		def _as_1d_array(name: str, value: arrayLike) -> npt.NDArray[np.floating]:

Uh oh!

Conversation

eduardojsbarroso commented Jun 19, 2026

Uh oh!

codecov Bot commented Jun 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

codecov Bot commented Jun 19, 2026 •

edited

Loading