def test_axis(self):
"""Test sigma calculation when axis is not None"""
self.mean, self.sigma = sigmaclip.calcsigma(self.data2d, self.errors2d, errors_as_weight=True)
self.assertAlmostEqual(self.mean, 2.75)
self.assertAlmostEqual(self.sigma, 1.898753053)
self.mean, self.sigma = sigmaclip.calcsigma(self.data2d, self.errors2d, axis=0, errors_as_weight=True)
self.assertEqual((abs(self.mean - numpy.array([2.0, 3.0])) < 1e-7).all(), True)
self.assertEqual((abs(self.sigma - numpy.array([2.0, 2.0])) < 1e-7).all(), True)
self.mean, self.sigma = sigmaclip.calcsigma(self.data2d, self.errors2d, axis=1, errors_as_weight=True)
self.assertEqual((abs(self.mean - numpy.array([0.75, 2.75, 4.75])) < 1e-7).all(), True)
self.assertEqual((abs(self.sigma - numpy.array(
[0.70710678, 0.70710678, 0.70710678])) < 1e-7).all(), True)
self.mean, self.sigma = sigmaclip.calcsigma(self.data3d, self.errors3d, axis=None, errors_as_weight=True)
self.assertAlmostEqual(self.mean, 11.75)
self.assertAlmostEqual(self.sigma, 7.10517533095)
self.mean, self.sigma = sigmaclip.calcsigma(self.data3d, self.errors3d, axis=0, errors_as_weight=True)
mean = numpy.array([[6., 7., 8., 9.], [10., 11., 12., 13.], [14., 15., 16., 17.]])
sigma = numpy.array([[8.48528137, 8.48528137, 8.48528137, 8.48528137],
[8.48528137, 8.48528137, 8.48528137, 8.48528137],
[8.48528137, 8.48528137, 8.48528137, 8.48528137]])
self.assertEqual((abs(self.mean - mean) < 1e-7).all(), True)
self.assertEqual((abs(self.sigma - sigma) < 1e-7).all(), True)
self.mean, self.sigma = sigmaclip.calcsigma(self.data3d, self.errors3d, axis=1, errors_as_weight=True)
mean = numpy.array([[4., 5., 6., 7.], [16., 17., 18., 19.]])
sigma = numpy.array([[4., 4., 4., 4.], [4., 4., 4., 4.]])
self.assertEqual((abs(self.mean - mean) < 1e-7).all(), True)
self.assertEqual((abs(self.sigma - sigma) < 1e-7).all(), True)
self.mean, self.sigma = sigmaclip.calcsigma(self.data3d, self.errors3d, axis=2, errors_as_weight=True)
mean = numpy.array([[1.75, 5.75, 9.75], [13.75, 17.75, 21.75]])
sigma = numpy.array([[1.31425748, 1.31425748, 1.31425748], [1.31425748, 1.31425748, 1.31425748]])
self.assertEqual((abs(self.mean - mean) < 1e-7).all(), True)
self.assertEqual((abs(self.sigma - sigma) < 1e-7).all(), True)
def calc_background(self, niter=-50, kappa=(5, 5)):
"""
Estimate background flux
Kwargs:
niter (int): number of iterations. Passed on to sigmaclip()
kappa (2-tuple of floats): lower and upper kappa
values. Passed on to sigmaclip()
Returns:
(dict): mean, sigma, indices
where light curve is at background level (True) and where
not (False).
Uses sigmaclipping to estimate a background. This only works
well when there are enough background points.
Also estimates the first point in time where the light curve
deviates from the background (T_zero), and the current duration
where the light curve is above the background.
"""
logger = logging.getLogger('tkp')
indices = numpy.ones(self.fluxes.shape, dtype=numpy.bool)
nniter = -niter if niter < 0 else niter
for i in range(nniter):
value = numpy.median(self.fluxes[indices])
# Get the sigma from the measured flux errors, instead of
# deriving it from the spread in values
sigma = numpy.mean(self.errors[indices])
# Throw away all data that are kappa*sigma above the current
# background value
newindices = numpy.logical_and(
self.fluxes < value + kappa[1] * sigma, indices)
if niter < 0:
if (newindices == indices).all(): # no change anymore
break
indices = newindices
# Now check if there are still data below the background
# Above, we have assumed most transients rise above the
# background, so we filter out increases in flux, not decreases
# We do that here now
# We can't do it at the same time as filtering the increase,
# because that may filter too much at one
for i in range(nniter):
value = numpy.median(self.fluxes[indices])
# Get the sigma from the measured flux errors, instead of
# deriving it from the spread in values
sigma = numpy.mean(self.errors[indices])
# Throw away all data that are kappa*sigma above the current
# background value
newindices = numpy.logical_and(
self.fluxes > value - kappa[0] * sigma, indices)
if niter < 0:
if (newindices == indices).all(): # no change anymore
break
indices = newindices
if len(numpy.where(indices)[0]) > 1:
value, sigma = calcsigma(self.fluxes[indices],
self.errors[indices])
# Now that we have the proper background, recalculate the indices
indices = (self.fluxes > value - sigma*kappa[0]) & (self.fluxes < value + sigma*kappa[1])
self.background['mean'] = value
self.background['sigma'] = sigma
self.background['indices'] = indices
return self.background
def test_weighted(self):
"""Calculate weighted mean and sample standard deviation"""
self.mean, self.sigma = sigmaclip.calcsigma(data=self.data, errors=self.errors, mean=None)
self.assertAlmostEqual(self.mean, 22.3759213759)
self.assertAlmostEqual(self.sigma, 1.15495684937)
def test_unweighted(self):
"""Calculate unweighted mean and sample standard deviation"""
self.mean, self.sigma = sigmaclip.calcsigma(data=self.data, errors=None, mean=None)
self.assertAlmostEqual(self.mean, 23.5714285714)
self.assertAlmostEqual(self.sigma, 3.40867241299)