Coverage for python/lsst/analysis/tools/actions/vector/calcRhoStatistics.py: 41%
110 statements
« prev ^ index » next coverage.py v7.5.0, created at 2024-05-01 04:53 -0700
« prev ^ index » next coverage.py v7.5.0, created at 2024-05-01 04:53 -0700
1# This file is part of analysis_tools.
2#
3# Developed for the LSST Data Management System.
4# This product includes software developed by the LSST Project
5# (https://www.lsst.org).
6# See the COPYRIGHT file at the top-level directory of this distribution
7# for details of code ownership.
8#
9# This program is free software: you can redistribute it and/or modify
10# it under the terms of the GNU General Public License as published by
11# the Free Software Foundation, either version 3 of the License, or
12# (at your option) any later version.
13#
14# This program is distributed in the hope that it will be useful,
15# but WITHOUT ANY WARRANTY; without even the implied warranty of
16# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17# GNU General Public License for more details.
18#
19# You should have received a copy of the GNU General Public License
20# along with this program. If not, see <https://www.gnu.org/licenses/>.
22from __future__ import annotations
24__all__ = (
25 "CalcRhoStatistics",
26 "TreecorrConfig",
27)
29import logging
30from typing import TYPE_CHECKING, Any, Mapping, cast
32import numpy as np
33import treecorr # type: ignore[import]
34from lsst.pex.config import ChoiceField, Config, ConfigField, Field, FieldValidationError
36from ...interfaces import KeyedData, KeyedDataAction, Vector
37from .calcMomentSize import CalcMomentSize
38from .ellipticity import CalcE, CalcEDiff
39from .mathActions import FractionalDifference
41if TYPE_CHECKING: 41 ↛ 42line 41 didn't jump to line 42, because the condition on line 41 was never true
42 from treecorr import GGCorrelation, KKCorrelation
44 from ...interfaces import KeyedDataSchema
46_LOG = logging.getLogger(__name__)
49class TreecorrConfig(Config):
50 """A Config class that holds some of the parameters supported by treecorr.
52 The fields in this class correspond to the parameters that can be passed to
53 any calls to `treecorr` methods, including catalog creation and two-point
54 correlation function calculations. The default values set for the fields
55 are identical to the default values set in v4.2 of `treecorr`.
57 A separate config class is used instead
58 of constructing a `~lsst.pex.config.DictField` so that mixed types can be
59 supported and the config can be validated at the beginning of the
60 execution.
62 Notes
63 -----
64 This is intended to be used in CalcRhoStatistics class. It only supports
65 some of the fields that are relevant for rho-statistics calculations.
66 """
68 nbins = Field[int]( 68 ↛ exitline 68 didn't jump to the function exit
69 doc=(
70 "How many bins to use. "
71 "(Exactly three of nbins, bin_size, min_sep, max_sep "
72 "are required. If nbins is not given, it will be "
73 "calculated from the values of the other three, "
74 "rounding up to the next highest integer. "
75 "In this case, bin_size will be readjusted to account "
76 "for this rounding up."
77 ),
78 optional=True,
79 check=lambda x: x > 0,
80 )
82 bin_size = Field[float](
83 doc=(
84 "The width of the bins in log(separation). "
85 "Exactly three of nbins, bin_size, min_sep, max_sep are required. "
86 "If bin_size is not given, it will be calculated from the values "
87 "of the other three."
88 ),
89 optional=True,
90 )
92 min_sep = Field[float](
93 doc=(
94 "The minimum separation in units of sep_units, if relevant. "
95 "Exactly three of nbins, bin_size, min_sep, max_sep are required. "
96 "If min_sep is not given, it will be calculated from the values "
97 "of the other three."
98 ),
99 optional=True,
100 )
102 max_sep = Field[float](
103 doc=(
104 "The maximum separation in units of sep_units, if relevant. "
105 "Exactly three of nbins, bin_size, min_sep, max_sep are required. "
106 "If max_sep is not given, it will be calculated from the values "
107 "of the other three."
108 ),
109 optional=True,
110 )
112 sep_units = ChoiceField[str](
113 doc=(
114 "The units to use for the separation values, given as a string. "
115 "This includes both min_sep and max_sep above, as well as the "
116 "units of the output distance values."
117 ),
118 default="radian",
119 optional=True,
120 allowed={units: units for units in ["arcsec", "arcmin", "degree", "hour", "radian"]},
121 )
123 bin_slop = Field[float](
124 doc=(
125 "How much slop to allow in the placement of pairs in the bins. "
126 "If bin_slop = 1, then the bin into which a particular pair is "
127 "placed may be incorrect by at most 1.0 bin widths. "
128 r"If None, use a bin_slop that gives a maximum error of 10% on "
129 "any bin, which has been found to yield good results for most "
130 "applications."
131 ),
132 default=None,
133 optional=True,
134 )
136 precision = Field[int]( 136 ↛ exitline 136 didn't jump to the function exit
137 doc=("The precision to use for the output values. This specifies how many digits to write."),
138 default=4,
139 optional=True,
140 check=lambda x: x > 0,
141 )
143 metric = ChoiceField[str](
144 doc=(
145 "Which metric to use for distance measurements. For details, see "
146 "https://rmjarvis.github.io/TreeCorr/_build/html/metric.html"
147 ),
148 default="Euclidean",
149 optional=True,
150 allowed={
151 "Euclidean": "straight-line Euclidean distance between two points",
152 "FisherRperp": (
153 "the perpendicular component of the distance, "
154 "following the definitions in "
155 "Fisher et al, 1994 (MNRAS, 267, 927)"
156 ),
157 "OldRperp": (
158 "the perpendicular component of the distance using the "
159 "definition of Rperp from TreeCorr v3.x."
160 ),
161 "Rlens": (
162 "Distance from the first object (taken to be a lens) to "
163 "the line connecting Earth and the second object "
164 "(taken to be a lensed source)."
165 ),
166 "Arc": "the true great circle distance for spherical coordinates.",
167 "Periodic": "Like ``Euclidean``, but with periodic boundaries.",
168 },
169 )
171 bin_type = ChoiceField[str](
172 doc="What type of binning should be used?",
173 default="Log",
174 optional=True,
175 allowed={
176 "Log": (
177 "Logarithmic binning in the distance. The bin steps will "
178 "be uniform in log(r) from log(min_sep) .. log(max_sep)."
179 ),
180 "Linear": (
181 "Linear binning in the distance. The bin steps will be "
182 "uniform in r from min_sep .. max_sep."
183 ),
184 "TwoD": (
185 "2-dimensional binning from x = (-max_sep .. max_sep) "
186 "and y = (-max_sep .. max_sep). The bin steps will be "
187 "uniform in both x and y. (i.e. linear in x,y)"
188 ),
189 },
190 )
192 var_method = ChoiceField[str](
193 doc="Which method to use for estimating the variance",
194 default="shot",
195 optional=True,
196 allowed={
197 method: method
198 for method in [
199 "shot",
200 "jackknife",
201 "sample",
202 "bootstrap",
203 "marked_bootstrap",
204 ]
205 },
206 )
208 npatch = Field[int](
209 doc="How many patches to split the catalog into for the purpose of "
210 "jackknife variance or other options that involve running via "
211 "patches (boostrap, marked_boostrap etc.)",
212 default=1,
213 optional=True,
214 )
216 num_bootstrap = Field[int](
217 doc=("How many bootstrap samples to use for the 'bootstrap' and 'marked_bootstrap' var methods."),
218 default=500,
219 optional=True,
220 )
222 rng_seed = Field[int](
223 doc="Value to seed the treecorr random number generator with. Used to generate patches.",
224 default=13579,
225 )
227 def validate(self):
228 # Docs inherited from base class
229 super().validate()
230 req_params = (self.nbins, self.bin_size, self.min_sep, self.max_sep)
231 num_req_params = sum(param is not None for param in req_params)
232 if num_req_params != 3:
233 msg = (
234 "You must specify exactly three of ``nbins``, ``bin_size``, ``min_sep`` and ``max_sep``"
235 f" in treecorr_config. {num_req_params} parameters were set instead."
236 )
237 raise FieldValidationError(self.__class__.bin_size, self, msg)
239 if self.min_sep is not None and self.max_sep is not None:
240 if self.min_sep > self.max_sep:
241 raise FieldValidationError(self.__class__.min_sep, self, "min_sep must be <= max_sep")
244class CalcRhoStatistics(KeyedDataAction):
245 r"""Calculate rho statistics.
247 Rho statistics refer to a collection of correlation functions involving
248 PSF ellipticity and size residuals. They quantify the contribution from PSF
249 leakage due to errors in PSF modeling to the weak lensing shear correlation
250 functions.
252 .. _rho_definitions:
254 The exact definitions of rho statistics as defined in [1]_ are given below.
256 .. math::
258 \rho_1(\theta) &= \left\langle
259 \delta e^*_{PSF}(x)
260 \delta e_{PSF}(x+\theta)
261 \right\rangle
263 \rho_2(\theta) &= \left\langle
264 e^*_{PSF}(x)
265 \delta e_{PSF}(x+\theta
266 \right\rangle
268 \rho_3(\theta) &= \left\langle
269 (e^*_{PSF}\frac{\delta T_{PSF}}{T_{PSF}}(x))
270 (e_{PSF}\frac{\delta T_{PSF}}{T_{PSF}})(x+\theta)
271 \right\rangle
273 \rho_4(\theta) &= \left\langle
274 \delta e^*_{PSF}(x)
275 (e_{PSF}\frac{\delta T_{PSF}}{T_{PSF}})(x+\theta)
276 \right\rangle
278 \rho_5(\theta) &= \left\langle
279 e^*_{PSF}(x)
280 (e_{PSF}\frac{\delta T_{PSF}}{T_{PSF}})(x+\theta)
281 \right\rangle
284 In addition to these five, we also compute the auto-correlation function of
285 the fractional size residuals and call it as the :math:`\rho'_3( \theta )`,
286 as referred to in Melchior et al. (2015) [2]_.
288 .. math::
290 \rho'_3(\theta) = \left\langle\frac{\delta T_{PSF}}{T_{PSF}}(x)
291 \frac{\delta T_{PSF}}{T_{PSF}}(x+\theta)
292 \right\rangle
295 The definition of ellipticity used in [1]_ correspond to shear-type,
296 which is typically smaller by a factor of 4 than using distortion-type.
298 References
299 ----------
301 .. [1] Jarvis, M., Sheldon, E., Zuntz, J., Kacprzak, T., Bridle, S. L.,
302 et. al (2016).
303 The DES Science Verification weak lensing shear catalogues
304 MNRAS, 460, 2245–2281.
305 https://doi.org/10.1093/mnras/stw990;
306 https://arxiv.org/abs/1507.05603
307 .. [2] Melchior, P., et. al (2015)
308 Mass and galaxy distributions of four massive galaxy clusters from
309 Dark Energy Survey Science Verification data
310 MNRAS, 449, no. 3, pp. 2219–2238.
311 https://doi:10.1093/mnras/stv398
312 https://arxiv.org/abs/1405.4285
313 """
315 colRa = Field[str](doc="RA column", default="coord_ra")
317 colDec = Field[str](doc="Dec column", default="coord_dec")
319 colXx = Field[str](doc="The column name to get the xx shape component from.", default="{band}_ixx")
321 colYy = Field[str](doc="The column name to get the yy shape component from.", default="{band}_iyy")
323 colXy = Field[str](doc="The column name to get the xy shape component from.", default="{band}_ixy")
325 colPsfXx = Field[str](
326 doc="The column name to get the PSF xx shape component from.", default="{band}_ixxPSF"
327 )
329 colPsfYy = Field[str](
330 doc="The column name to get the PSF yy shape component from.", default="{band}_iyyPSF"
331 )
333 colPsfXy = Field[str](
334 doc="The column name to get the PSF xy shape component from.", default="{band}_ixyPSF"
335 )
337 ellipticityType = ChoiceField[str](
338 doc="The type of ellipticity to calculate",
339 optional=False,
340 allowed={
341 "distortion": r"Distortion, measured as :math:`(I_{xx}-I_{yy})/(I_{xx}+I_{yy})`",
342 "shear": (
343 r"Shear, measured as :math:`(I_{xx}-I_{yy})/(I_{xx}+I_{yy}+2\sqrt{I_{xx}I_{yy}-I_{xy}^2})`"
344 ),
345 },
346 default="distortion",
347 )
349 sizeType = ChoiceField[str](
350 doc="The type of size to calculate",
351 default="trace",
352 allowed={
353 "trace": "trace radius",
354 "determinant": "determinant radius",
355 },
356 )
358 treecorr = ConfigField[TreecorrConfig](
359 doc="TreeCorr configuration",
360 )
362 def setDefaults(self):
363 super().setDefaults()
364 self.treecorr = TreecorrConfig()
365 self.treecorr.sep_units = "arcmin"
366 self.treecorr.max_sep = 100.0
368 def getInputSchema(self) -> KeyedDataSchema:
369 return (
370 (self.colRa, Vector),
371 (self.colDec, Vector),
372 (self.colXx, Vector),
373 (self.colYy, Vector),
374 (self.colXy, Vector),
375 (self.colPsfXx, Vector),
376 (self.colPsfYy, Vector),
377 (self.colPsfXy, Vector),
378 )
380 def __call__(self, data: KeyedData, **kwargs) -> KeyedData:
381 calcEMeas = CalcE(
382 colXx=self.colXx,
383 colYy=self.colYy,
384 colXy=self.colXy,
385 ellipticityType=self.ellipticityType,
386 )
387 calcEpsf = CalcE(
388 colXx=self.colPsfXx,
389 colYy=self.colPsfYy,
390 colXy=self.colPsfXy,
391 ellipticityType=self.ellipticityType,
392 )
394 calcEDiff = CalcEDiff(colA=calcEMeas, colB=calcEpsf)
396 calcSizeResidual = FractionalDifference(
397 actionA=CalcMomentSize(
398 colXx=self.colXx,
399 colYy=self.colYy,
400 colXy=self.colXy,
401 sizeType=self.sizeType,
402 ),
403 actionB=CalcMomentSize(
404 colXx=self.colPsfXx,
405 colYy=self.colPsfYy,
406 colXy=self.colPsfXy,
407 sizeType=self.sizeType,
408 ),
409 )
411 # distortion-type ellipticity has a shear response of 2, so we need to
412 # divide by 2 so that the rho-stats do not depend on the
413 # ellipticity-type.
414 # Note: For distortion, the responsitivity is 2(1 - e^2_{rms}),
415 # where e_rms is the root mean square ellipticity per component.
416 # This is expected to be small and we ignore it here.
417 # This definition of responsitivity is consistent with the definions
418 # used in the rho-statistics calculations for the HSC shear catalog
419 # papers (Mandelbaum et al. 2018, Li et al., 2022).
420 responsitivity = 2.0 if self.ellipticityType == "distortion" else 1.0
422 # Call the actions on the data.
423 eMEAS = calcEMeas(data, **kwargs)
424 if self.ellipticityType == "distortion":
425 _LOG.debug("Correction value of responsitivity would be %f", 2 - np.mean(np.abs(eMEAS) ** 2))
426 eMEAS /= responsitivity # type: ignore
427 e1, e2 = np.real(eMEAS), np.imag(eMEAS)
428 eRes = calcEDiff(data, **kwargs)
429 eRes /= responsitivity # type: ignore
430 e1Res, e2Res = np.real(eRes), np.imag(eRes)
431 sizeRes = calcSizeResidual(data, **kwargs)
433 # Scale the sizeRes by ellipticities
434 e1SizeRes = e1 * sizeRes
435 e2SizeRes = e2 * sizeRes
437 # Package the arguments to capture auto-/cross-correlations for the
438 # Rho statistics.
439 args = {
440 0: (sizeRes, None),
441 1: (e1Res, e2Res, None, None),
442 2: (e1, e2, e1Res, e2Res),
443 3: (e1SizeRes, e2SizeRes, None, None),
444 4: (e1Res, e2Res, e1SizeRes, e2SizeRes),
445 5: (e1, e2, e1SizeRes, e2SizeRes),
446 }
448 ra: Vector = data[self.colRa] # type: ignore
449 dec: Vector = data[self.colDec] # type: ignore
451 treecorr_config_dict = self.treecorr.toDict()
453 # Swap rng_seed with an rng instance in treecorr config.
454 rng = np.random.RandomState(treecorr_config_dict.pop("rng_seed"))
455 treecorr_config_dict["rng"] = rng
457 # Pass the appropriate arguments to the correlator and build a dict
458 rhoStats: Mapping[str, treecorr.BinnedCorr2] = {}
459 for rhoIndex in range(1, 6):
460 _LOG.info("Calculating rho-%d", rhoIndex)
461 rhoStats[f"rho{rhoIndex}"] = self._corrSpin2( # type: ignore[index]
462 ra,
463 dec,
464 *(args[rhoIndex]),
465 treecorr_config_dict=treecorr_config_dict,
466 )
468 _LOG.info("Calculating rho3alt")
469 rhoStats["rho3alt"] = self._corrSpin0( # type: ignore[index]
470 ra,
471 dec,
472 *(args[0]),
473 treecorr_config_dict=treecorr_config_dict,
474 )
475 return cast(KeyedData, rhoStats)
477 @classmethod
478 def _corrSpin0(
479 cls,
480 ra: Vector,
481 dec: Vector,
482 k1: Vector,
483 k2: Vector | None = None,
484 raUnits: str = "degrees",
485 decUnits: str = "degrees",
486 treecorr_config_dict: Mapping[str, Any] | None = None,
487 ) -> KKCorrelation:
488 """Function to compute correlations between at most two scalar fields.
490 This is used to compute rho3alt statistics, given the appropriate
491 spin-0 (scalar) fields, usually fractional size residuals.
493 Parameters
494 ----------
495 ra : `numpy.array`
496 The right ascension values of entries in the catalog.
497 dec : `numpy.array`
498 The declination values of entries in the catalog.
499 k1 : `numpy.array`
500 The primary scalar field.
501 k2 : `numpy.array`, optional
502 The secondary scalar field.
503 Autocorrelation of the primary field is computed if `None`.
504 raUnits : `str`, optional
505 Unit of the right ascension values. Valid options are
506 "degrees", "arcmin", "arcsec", "hours" or "radians".
507 decUnits : `str`, optional
508 Unit of the declination values. Valid options are
509 "degrees", "arcmin", "arcsec", "hours" or "radians".
510 treecorr_config_dict: `dict`, optional
511 Config dictionary to be passed to `treecorr`
512 (`treecorr.KKCorrelation` or `treecorr.Catalog`).
514 Returns
515 -------
516 xy : `treecorr.KKCorrelation`
517 A `treecorr.KKCorrelation` object containing the correlation
518 function.
519 """
520 _LOG.debug(
521 "No. of entries: %d. The number of pairs in the resulting KKCorrelation cannot exceed %d",
522 len(ra),
523 len(ra) * (len(ra) - 1) / 2,
524 )
525 xy = treecorr.KKCorrelation(config=treecorr_config_dict)
526 catA = treecorr.Catalog(
527 config=treecorr_config_dict,
528 ra=ra,
529 dec=dec,
530 k=k1,
531 ra_units=raUnits,
532 dec_units=decUnits,
533 logger=_LOG,
534 )
535 if k2 is None:
536 # Calculate the auto-correlation
537 xy.process(catA)
538 else:
539 catB = treecorr.Catalog(
540 config=treecorr_config_dict,
541 ra=ra,
542 dec=dec,
543 k=k2,
544 ra_units=raUnits,
545 dec_units=decUnits,
546 logger=_LOG,
547 patch_centers=catA.patch_centers,
548 )
549 # Calculate the cross-correlation
550 xy.process(catA, catB)
552 _LOG.debug("Correlated %d pairs based on the config set.", sum(xy.npairs))
553 return xy
555 @classmethod
556 def _corrSpin2(
557 cls,
558 ra: Vector,
559 dec: Vector,
560 g1a: Vector,
561 g2a: Vector,
562 g1b: Vector | None = None,
563 g2b: Vector | None = None,
564 raUnits: str = "degrees",
565 decUnits: str = "degrees",
566 treecorr_config_dict: Mapping[str, Any] | None = None,
567 ) -> GGCorrelation:
568 """Function to compute correlations between shear-like fields.
570 This is used to compute Rho statistics, given the appropriate spin-2
571 (shear-like) fields.
573 Parameters
574 ----------
575 ra : `numpy.array`
576 The right ascension values of entries in the catalog.
577 dec : `numpy.array`
578 The declination values of entries in the catalog.
579 g1a : `numpy.array`
580 The first component of the primary shear-like field.
581 g2a : `numpy.array`
582 The second component of the primary shear-like field.
583 g1b : `numpy.array`, optional
584 The first component of the secondary shear-like field.
585 Autocorrelation of the primary field is computed if `None`.
586 g2b : `numpy.array`, optional
587 The second component of the secondary shear-like field.
588 Autocorrelation of the primary field is computed if `None`.
589 raUnits : `str`, optional
590 Unit of the right ascension values. Valid options are
591 "degrees", "arcmin", "arcsec", "hours" or "radians".
592 decUnits : `str`, optional
593 Unit of the declination values. Valid options are
594 "degrees", "arcmin", "arcsec", "hours" or "radians".
595 treecorr_config_dict : `dict`, optional
596 Config dictionary to be passed to `treecorr`
597 (`treecorr.GGCorrelation` or `treecorr.Catalog`).
599 Returns
600 -------
601 xy : `treecorr.GGCorrelation`
602 A `treecorr.GGCorrelation` object containing the correlation
603 function.
604 """
605 _LOG.debug(
606 "No. of entries: %d. The number of pairs in the resulting GGCorrelation cannot exceed %d",
607 len(ra),
608 len(ra) * (len(ra) - 1) / 2,
609 )
610 xy = treecorr.GGCorrelation(config=treecorr_config_dict)
611 catA = treecorr.Catalog(
612 config=treecorr_config_dict,
613 ra=ra,
614 dec=dec,
615 g1=g1a,
616 g2=g2a,
617 ra_units=raUnits,
618 dec_units=decUnits,
619 logger=_LOG,
620 )
621 if g1b is None or g2b is None:
622 # Calculate the auto-correlation
623 xy.process(catA)
624 else:
625 catB = treecorr.Catalog(
626 config=treecorr_config_dict,
627 ra=ra,
628 dec=dec,
629 g1=g1b,
630 g2=g2b,
631 ra_units=raUnits,
632 dec_units=decUnits,
633 logger=_LOG,
634 patch_centers=catA.patch_centers,
635 )
636 # Calculate the cross-correlation
637 xy.process(catA, catB)
639 _LOG.debug("Correlated %d pairs based on the config set.", sum(xy.npairs))
640 return xy