Coverage for python/lsst/analysis/tools/actions/vector/calcRhoStatistics.py: 39%

106 statements  

« prev     ^ index     » next       coverage.py v7.2.5, created at 2023-05-02 11:54 -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/>. 

21 

22from __future__ import annotations 

23 

24__all__ = ( 

25 "BinnedCorr2Config", 

26 "CalcRhoStatistics", 

27) 

28 

29import logging 

30from typing import TYPE_CHECKING, Any, Mapping, cast 

31 

32import numpy as np 

33import treecorr # type: ignore[import] 

34from lsst.pex.config import ChoiceField, Config, ConfigField, Field, FieldValidationError 

35 

36from ...interfaces import KeyedData, KeyedDataAction, Vector 

37from .calcShapeSize import CalcShapeSize 

38from .ellipticity import CalcE, CalcEDiff 

39from .vectorActions import FractionalDifference 

40 

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 

43 

44 from ...interfaces import KeyedDataSchema 

45 

46_LOG = logging.getLogger(__name__) 

47 

48 

49class BinnedCorr2Config(Config): 

50 """A Config class that holds some of the parameters supported by treecorr. 

51 

52 The fields in this class correspond to the parameters that can be passed to 

53 BinnedCorr2 in `treecorr`, which is the base class for all 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`. The 

56 parameters that are excluded in this class are 

57 'verbose', 'log_file', 'output_dots', 'rng' and 'pairwise' (deprecated). 

58 For details about these options, see the documentation for `treecorr`: 

59 https://rmjarvis.github.io/TreeCorr/_build/html/correlation2.html 

60 

61 A separate config class is used instead 

62 of constructing a `~lsst.pex.config.DictField` so that mixed types can be 

63 supported and the config can be validated at the beginning of the 

64 execution. 

65 

66 Notes 

67 ----- 

68 This is intended to be used in CalcRhoStatistics class. It only supports 

69 some of the fields that are relevant for rho-statistics calculations. 

70 """ 

71 

72 nbins = Field[int]( 72 ↛ exitline 72 didn't jump to the function exit

73 doc=( 

74 "How many bins to use. " 

75 "(Exactly three of nbins, bin_size, min_sep, max_sep " 

76 "are required. If nbins is not given, it will be " 

77 "calculated from the values of the other three, " 

78 "rounding up to the next highest integer. " 

79 "In this case, bin_size will be readjusted to account " 

80 "for this rounding up." 

81 ), 

82 optional=True, 

83 check=lambda x: x > 0, 

84 ) 

85 

86 bin_size = Field[float]( 

87 doc=( 

88 "The width of the bins in log(separation). " 

89 "Exactly three of nbins, bin_size, min_sep, max_sep are required. " 

90 "If bin_size is not given, it will be calculated from the values " 

91 "of the other three." 

92 ), 

93 optional=True, 

94 ) 

95 

96 min_sep = Field[float]( 

97 doc=( 

98 "The minimum separation in units of sep_units, if relevant. " 

99 "Exactly three of nbins, bin_size, min_sep, max_sep are required. " 

100 "If min_sep is not given, it will be calculated from the values " 

101 "of the other three." 

102 ), 

103 optional=True, 

104 ) 

105 

106 max_sep = Field[float]( 

107 doc=( 

108 "The maximum separation in units of sep_units, if relevant. " 

109 "Exactly three of nbins, bin_size, min_sep, max_sep are required. " 

110 "If max_sep is not given, it will be calculated from the values " 

111 "of the other three." 

112 ), 

113 optional=True, 

114 ) 

115 

116 sep_units = ChoiceField[str]( 

117 doc=( 

118 "The units to use for the separation values, given as a string. " 

119 "This includes both min_sep and max_sep above, as well as the " 

120 "units of the output distance values." 

121 ), 

122 default="radian", 

123 optional=True, 

124 allowed={units: units for units in ["arcsec", "arcmin", "degree", "hour", "radian"]}, 

125 ) 

126 

127 bin_slop = Field[float]( 

128 doc=( 

129 "How much slop to allow in the placement of pairs in the bins. " 

130 "If bin_slop = 1, then the bin into which a particular pair is " 

131 "placed may be incorrect by at most 1.0 bin widths. " 

132 r"If None, use a bin_slop that gives a maximum error of 10% on " 

133 "any bin, which has been found to yield good results for most " 

134 "applications." 

135 ), 

136 default=None, 

137 optional=True, 

138 ) 

139 

140 precision = Field[int]( 140 ↛ exitline 140 didn't jump to the function exit

141 doc=("The precision to use for the output values. This specifies how many digits to write."), 

142 default=4, 

143 optional=True, 

144 check=lambda x: x > 0, 

145 ) 

146 

147 metric = ChoiceField[str]( 

148 doc=( 

149 "Which metric to use for distance measurements. For details, see " 

150 "https://rmjarvis.github.io/TreeCorr/_build/html/metric.html" 

151 ), 

152 default="Euclidean", 

153 optional=True, 

154 allowed={ 

155 "Euclidean": "straight-line Euclidean distance between two points", 

156 "FisherRperp": ( 

157 "the perpendicular component of the distance, " 

158 "following the definitions in " 

159 "Fisher et al, 1994 (MNRAS, 267, 927)" 

160 ), 

161 "OldRperp": ( 

162 "the perpendicular component of the distance using the " 

163 "definition of Rperp from TreeCorr v3.x." 

164 ), 

165 "Rlens": ( 

166 "Distance from the first object (taken to be a lens) to " 

167 "the line connecting Earth and the second object " 

168 "(taken to be a lensed source)." 

169 ), 

170 "Arc": "the true great circle distance for spherical coordinates.", 

171 "Periodic": "Like ``Euclidean``, but with periodic boundaries.", 

172 }, 

173 ) 

174 

175 bin_type = ChoiceField[str]( 

176 doc="What type of binning should be used?", 

177 default="Log", 

178 optional=True, 

179 allowed={ 

180 "Log": ( 

181 "Logarithmic binning in the distance. The bin steps will " 

182 "be uniform in log(r) from log(min_sep) .. log(max_sep)." 

183 ), 

184 "Linear": ( 

185 "Linear binning in the distance. The bin steps will be " 

186 "uniform in r from min_sep .. max_sep." 

187 ), 

188 "TwoD": ( 

189 "2-dimensional binning from x = (-max_sep .. max_sep) " 

190 "and y = (-max_sep .. max_sep). The bin steps will be " 

191 "uniform in both x and y. (i.e. linear in x,y)" 

192 ), 

193 }, 

194 ) 

195 

196 var_method = ChoiceField[str]( 

197 doc="Which method to use for estimating the variance", 

198 default="shot", 

199 optional=True, 

200 allowed={ 

201 method: method 

202 for method in [ 

203 "shot", 

204 "jackknife", 

205 "sample", 

206 "bootstrap", 

207 "marked_bootstrap", 

208 ] 

209 }, 

210 ) 

211 

212 num_bootstrap = Field[int]( 

213 doc=("How many bootstrap samples to use for the 'bootstrap' and 'marked_bootstrap' var methods."), 

214 default=500, 

215 optional=True, 

216 ) 

217 

218 def validate(self): 

219 # Docs inherited from base class 

220 super().validate() 

221 req_params = (self.nbins, self.bin_size, self.min_sep, self.max_sep) 

222 num_req_params = sum(param is not None for param in req_params) 

223 if num_req_params != 3: 

224 msg = ( 

225 "You must specify exactly three of ``nbins``, ``bin_size``, ``min_sep`` and ``max_sep``" 

226 f" in treecorr_config. {num_req_params} parameters were set instead." 

227 ) 

228 raise FieldValidationError(self.__class__.bin_size, self, msg) 

229 

230 if self.min_sep is not None and self.max_sep is not None: 

231 if self.min_sep > self.max_sep: 

232 raise FieldValidationError(self.__class__.min_sep, self, "min_sep must be <= max_sep") 

233 

234 

235class CalcRhoStatistics(KeyedDataAction): 

236 r"""Calculate rho statistics 

237 

238 Rho statistics refer to a collection of correlation functions involving 

239 PSF ellipticity and size residuals. They quantify the contribution from PSF 

240 leakage due to errors in PSF modeling to the weak lensing shear correlation 

241 functions. The standard rho statistics are indexed from 1 to 5, and 

242 this action calculates a sixth rho statistic, indexed 0. 

243 

244 Notes 

245 ----- 

246 The exact definitions of rho statistics as defined in [1]_ are given below. 

247 In addition to these five, we also compute the auto-correlation function of 

248 the fractional size residuals and call it as the :math:`\rho_0( \theta )`. 

249 

250 .. math:: 

251 

252 \rho_1(\theta) &= \langle \delta e^*_{PSF}(x) \delta e_{PSF}(x+\theta) \rangle # noqa: W505 

253 

254 \rho_2(\theta) &= \langle e^*_{PSF}(x) \delta e_{PSF}(x+\theta) \rangle 

255 

256 \rho_3(\theta) &= \left\langle (e^*_{PSF}\frac{\delta T_{PSF}}{T_{PSF}}(x)) 

257 \delta e_{PSF}(x+\theta) \right\rangle 

258 

259 \rho_4(\theta) &= \left\langle (\delta e^*_{PSF}(x) 

260 (e_{PSF}\frac{\delta T_{PSF}}{T_{PSF}}(x+\theta)) \right\rangle 

261 

262 \rho_5(\theta) &= \left\langle (e^*_{PSF}(x) 

263 (e_{PSF}\frac{\delta T_{PSF}}{T_{PSF}}(x+\theta)) \right\rangle 

264 

265 There is a slightly different version for :math:`\rho_3( \theta )`, used in Melchior et al. (2015) [2]_. 

266 

267 .. math:: 

268 

269 \rho'_3(\theta) &= \left\langle\frac{\delta T_{PSF}}{T_{PSF}}(x) 

270 \frac{\delta T_{PSF}}{T_{PSF}}(x+\theta) 

271 \right\rangle 

272 

273 

274 The definition of ellipticity used in [1]_ correspond to ``shear``-type ellipticity, which is typically 

275 smaller by a factor of 4 than using ``distortion``-type ellipticity. 

276 

277 References 

278 ---------- 

279 .. [1] Jarvis, M., Sheldon, E., Zuntz, J., Kacprzak, T., Bridle, S. L., et. al (2016). # noqa: W501 

280 The DES Science Verification weak lensing shear catalogues 

281 MNRAS, 460, 2245–2281. 

282 https://doi.org/10.1093/mnras/stw990; 

283 https://arxiv.org/abs/1507.05603 

284 .. [2] Melchior, P., et. al (2015) 

285 Mass and galaxy distributions of four massive galaxy clusters from Dark Energy Survey 

286 Science Verification data 

287 MNRAS, 449, no. 3, pp. 2219–2238. 

288 https://doi:10.1093/mnras/stv398 

289 https://arxiv.org/abs/1405.4285 

290 """ 

291 

292 colRa = Field[str](doc="RA column", default="coord_ra") 

293 

294 colDec = Field[str](doc="Dec column", default="coord_dec") 

295 

296 colXx = Field[str](doc="The column name to get the xx shape component from.", default="{band}_ixx") 

297 

298 colYy = Field[str](doc="The column name to get the yy shape component from.", default="{band}_iyy") 

299 

300 colXy = Field[str](doc="The column name to get the xy shape component from.", default="{band}_ixy") 

301 

302 colPsfXx = Field[str]( 

303 doc="The column name to get the PSF xx shape component from.", default="{band}_ixxPSF" 

304 ) 

305 

306 colPsfYy = Field[str]( 

307 doc="The column name to get the PSF yy shape component from.", default="{band}_iyyPSF" 

308 ) 

309 

310 colPsfXy = Field[str]( 

311 doc="The column name to get the PSF xy shape component from.", default="{band}_ixyPSF" 

312 ) 

313 

314 ellipticityType = ChoiceField[str]( 

315 doc="The type of ellipticity to calculate", 

316 allowed={ 

317 "distortion": "Distortion, measured as (Ixx - Iyy)/(Ixx + Iyy)", 

318 "shear": ("Shear, measured as (Ixx - Iyy)/(Ixx + Iyy + 2*sqrt(Ixx*Iyy - Ixy**2))"), 

319 }, 

320 default="distortion", 

321 ) 

322 

323 sizeType = ChoiceField[str]( 

324 doc="The type of size to calculate", 

325 default="trace", 

326 allowed={ 

327 "trace": "trace radius", 

328 "determinant": "determinant radius", 

329 }, 

330 ) 

331 

332 treecorr = ConfigField[BinnedCorr2Config]( 

333 doc="TreeCorr configuration", 

334 ) 

335 

336 def setDefaults(self): 

337 super().setDefaults() 

338 self.treecorr = BinnedCorr2Config() 

339 self.treecorr.sep_units = "arcmin" 

340 self.treecorr.max_sep = 100.0 

341 

342 def getInputSchema(self) -> KeyedDataSchema: 

343 return ( 

344 (self.colRa, Vector), 

345 (self.colDec, Vector), 

346 (self.colXx, Vector), 

347 (self.colYy, Vector), 

348 (self.colXy, Vector), 

349 (self.colPsfXx, Vector), 

350 (self.colPsfYy, Vector), 

351 (self.colPsfXy, Vector), 

352 ) 

353 

354 def __call__(self, data: KeyedData, **kwargs) -> KeyedData: 

355 calcEMeas = CalcE( 

356 colXx=self.colXx, 

357 colYy=self.colYy, 

358 colXy=self.colXy, 

359 ellipticityType=self.ellipticityType, 

360 ) 

361 calcEpsf = CalcE( 

362 colXx=self.colPsfXx, 

363 colYy=self.colPsfYy, 

364 colXy=self.colPsfXy, 

365 ellipticityType=self.ellipticityType, 

366 ) 

367 

368 calcEDiff = CalcEDiff(colA=calcEMeas, colB=calcEpsf) 

369 

370 calcSizeResidual = FractionalDifference( 

371 actionA=CalcShapeSize( 

372 colXx=self.colXx, 

373 colYy=self.colYy, 

374 colXy=self.colXy, 

375 sizeType=self.sizeType, 

376 ), 

377 actionB=CalcShapeSize( 

378 colXx=self.colPsfXx, 

379 colYy=self.colPsfYy, 

380 colXy=self.colPsfXy, 

381 sizeType=self.sizeType, 

382 ), 

383 ) 

384 

385 # distortion-type ellipticity has a shear response of 2, so we need to 

386 # divide by 2 so that the rho-stats do not depend on the 

387 # ellipticity-type. 

388 # Note: For distortion, the responsitivity is 2(1 - e^2_{rms}), 

389 # where e_rms is the root mean square ellipticity per component. 

390 # This is expected to be small and we ignore it here. 

391 # This definition of responsitivity is consistent with the definions 

392 # used in the rho-statistics calculations for the HSC shear catalog 

393 # papers (Mandelbaum et al. 2018, Li et al., 2022). 

394 responsitivity = 2.0 if self.ellipticityType == "distortion" else 1.0 

395 

396 # Call the actions on the data. 

397 eMEAS = calcEMeas(data, **kwargs) 

398 if self.ellipticityType == "distortion": 

399 _LOG.debug("Correction value of responsitivity would be %f", 2 - np.mean(np.abs(eMEAS) ** 2)) 

400 eMEAS /= responsitivity # type: ignore 

401 e1, e2 = np.real(eMEAS), np.imag(eMEAS) 

402 eRes = calcEDiff(data, **kwargs) 

403 eRes /= responsitivity # type: ignore 

404 e1Res, e2Res = np.real(eRes), np.imag(eRes) 

405 sizeRes = calcSizeResidual(data, **kwargs) 

406 

407 # Scale the sizeRes by ellipticities 

408 e1SizeRes = e1 * sizeRes 

409 e2SizeRes = e2 * sizeRes 

410 

411 # Package the arguments to capture auto-/cross-correlations for the 

412 # Rho statistics. 

413 args = { 

414 0: (sizeRes, None), 

415 1: (e1Res, e2Res, None, None), 

416 2: (e1, e2, e1Res, e2Res), 

417 3: (e1SizeRes, e2SizeRes, None, None), 

418 4: (e1Res, e2Res, e1SizeRes, e2SizeRes), 

419 5: (e1, e2, e1SizeRes, e2SizeRes), 

420 } 

421 

422 ra: Vector = data[self.colRa] # type: ignore 

423 dec: Vector = data[self.colDec] # type: ignore 

424 

425 treecorrKwargs = self.treecorr.toDict() 

426 

427 # Pass the appropriate arguments to the correlator and build a dict 

428 rhoStats: Mapping[str, treecorr.BinnedCorr2] = {} 

429 for rhoIndex in range(1, 6): 

430 _LOG.info("Calculating rho-%d", rhoIndex) 

431 rhoStats[f"rho{rhoIndex}"] = self._corrSpin2( # type: ignore[index] 

432 ra, dec, *(args[rhoIndex]), **treecorrKwargs 

433 ) 

434 

435 _LOG.info("Calculating rho3alt") 

436 rhoStats["rho3alt"] = self._corrSpin0(ra, dec, *(args[0]), **treecorrKwargs) # type: ignore[index] 

437 return cast(KeyedData, rhoStats) 

438 

439 @classmethod 

440 def _corrSpin0( 

441 cls, 

442 ra: Vector, 

443 dec: Vector, 

444 k1: Vector, 

445 k2: Vector | None = None, 

446 raUnits: str = "degrees", 

447 decUnits: str = "degrees", 

448 **treecorrKwargs: Any, 

449 ) -> KKCorrelation: 

450 """Function to compute correlations between at most two scalar fields. 

451 

452 This is used to compute rho3alt statistics, given the appropriate 

453 spin-0 (scalar) fields, usually fractional size residuals. 

454 

455 Parameters 

456 ---------- 

457 ra : `numpy.array` 

458 The right ascension values of entries in the catalog. 

459 dec : `numpy.array` 

460 The declination values of entries in the catalog. 

461 k1 : `numpy.array` 

462 The primary scalar field. 

463 k2 : `numpy.array`, optional 

464 The secondary scalar field. 

465 Autocorrelation of the primary field is computed if `None`. 

466 raUnits : `str`, optional 

467 Unit of the right ascension values. Valid options are 

468 "degrees", "arcmin", "arcsec", "hours" or "radians". 

469 decUnits : `str`, optional 

470 Unit of the declination values. Valid options are 

471 "degrees", "arcmin", "arcsec", "hours" or "radians". 

472 **treecorrKwargs 

473 Keyword arguments to be passed to `treecorr.KKCorrelation`. 

474 

475 Returns 

476 ------- 

477 xy : `treecorr.KKCorrelation` 

478 A `treecorr.KKCorrelation` object containing the correlation 

479 function. 

480 """ 

481 _LOG.debug( 

482 "No. of entries: %d. The number of pairs in the resulting KKCorrelation cannot exceed %d", 

483 len(ra), 

484 len(ra) * (len(ra) - 1) / 2, 

485 ) 

486 xy = treecorr.KKCorrelation(**treecorrKwargs) 

487 catA = treecorr.Catalog(ra=ra, dec=dec, k=k1, ra_units=raUnits, dec_units=decUnits, logger=_LOG) 

488 if k2 is None: 

489 # Calculate the auto-correlation 

490 xy.process(catA) 

491 else: 

492 catB = treecorr.Catalog(ra=ra, dec=dec, k=k2, ra_units=raUnits, dec_units=decUnits, logger=_LOG) 

493 # Calculate the cross-correlation 

494 xy.process(catA, catB) 

495 

496 _LOG.debug("Correlated %d pairs based on the config set.", sum(xy.npairs)) 

497 return xy 

498 

499 @classmethod 

500 def _corrSpin2( 

501 cls, 

502 ra: Vector, 

503 dec: Vector, 

504 g1a: Vector, 

505 g2a: Vector, 

506 g1b: Vector | None = None, 

507 g2b: Vector | None = None, 

508 raUnits: str = "degrees", 

509 decUnits: str = "degrees", 

510 **treecorrKwargs: Any, 

511 ) -> GGCorrelation: 

512 """Function to compute correlations between shear-like fields. 

513 

514 This is used to compute Rho statistics, given the appropriate spin-2 

515 (shear-like) fields. 

516 

517 Parameters 

518 ---------- 

519 ra : `numpy.array` 

520 The right ascension values of entries in the catalog. 

521 dec : `numpy.array` 

522 The declination values of entries in the catalog. 

523 g1a : `numpy.array` 

524 The first component of the primary shear-like field. 

525 g2a : `numpy.array` 

526 The second component of the primary shear-like field. 

527 g1b : `numpy.array`, optional 

528 The first component of the secondary shear-like field. 

529 Autocorrelation of the primary field is computed if `None`. 

530 g2b : `numpy.array`, optional 

531 The second component of the secondary shear-like field. 

532 Autocorrelation of the primary field is computed if `None`. 

533 raUnits : `str`, optional 

534 Unit of the right ascension values. Valid options are 

535 "degrees", "arcmin", "arcsec", "hours" or "radians". 

536 decUnits : `str`, optional 

537 Unit of the declination values. Valid options are 

538 "degrees", "arcmin", "arcsec", "hours" or "radians". 

539 **treecorrKwargs 

540 Keyword arguments to be passed to `treecorr.GGCorrelation`. 

541 

542 Returns 

543 ------- 

544 xy : `treecorr.GGCorrelation` 

545 A `treecorr.GGCorrelation` object containing the correlation 

546 function. 

547 """ 

548 _LOG.debug( 

549 "No. of entries: %d. The number of pairs in the resulting GGCorrelation cannot exceed %d", 

550 len(ra), 

551 len(ra) * (len(ra) - 1) / 2, 

552 ) 

553 xy = treecorr.GGCorrelation(**treecorrKwargs) 

554 catA = treecorr.Catalog( 

555 ra=ra, dec=dec, g1=g1a, g2=g2a, ra_units=raUnits, dec_units=decUnits, logger=_LOG 

556 ) 

557 if g1b is None or g2b is None: 

558 # Calculate the auto-correlation 

559 xy.process(catA) 

560 else: 

561 catB = treecorr.Catalog( 

562 ra=ra, dec=dec, g1=g1b, g2=g2b, ra_units=raUnits, dec_units=decUnits, logger=_LOG 

563 ) 

564 # Calculate the cross-correlation 

565 xy.process(catA, catB) 

566 

567 _LOG.debug("Correlated %d pairs based on the config set.", sum(xy.npairs)) 

568 return xy