def NZAD(self, vmin = 0.005, vmax = 1.995, delv = 0.05): """ Compute the Anderson Darling statistic and p-value for the two distributions of sumpz and true_z vector of spec-z's Since the Anderson Darling test requires a properly normalized distribution over the [vmin,vmax] range, will need to create a new qp object defined on the range np.arange(vmin,vmax+delv,delv) Parameters: vmin, vmax: specz values outside of these values are discarded delz: grid spacing for [vmin,vmax] interval to create new qp object ----------- using: string which parameterization to evaluate Returns: -------- Anderson-Darling statistic and pvalue """ #copy the form of Rongpu's use of skgof functions #will have to use QPPDFCDF class, as those expect objects #that have a .cdf method for a vector of values print "using %f and %f for vmin and vmax\n"%(vmin,vmax) szs = self.truth mask = (szs > vmin) & (szs < vmax) vgrid = np.arange(vmin,vmax+delv,delv) veval = self.stackpz.evaluate(vgrid,'gridded',True,False) vobj = qp.PDF(gridded = (veval[0],veval[1])) tmpnzfunc = QPPDFCDF(vobj,self.dx) nzAD = skgof.ad_test(szs[mask],tmpnzfunc) return nzAD.statistic, nzAD.pvalue
def AD(self, using, dx=0.0001, vmin=0.005, vmax=0.995):
"""
Compute the Anderson-Darling statistic and p-value for the PIT
values by comparing with a uniform distribution between 0 and 1.
Since the statistic diverges at 0 and 1, PIT values too close to
0 or 1 are discarded.
Parameters:
-----------
using: string
which parameterization to evaluate
dx: float
step size for integral
vmin, vmax: floats
PIT values outside this range are discarded
Returns:
--------
AD statistic and pvalue
"""
if self.pitarray is not None:
pits = np.array(self.pitarray)
else:
pits = np.array(self.PIT(using=using,dx=dx))
self.pitarray = pits
mask = (pits>vmin) & (pits<vmax)
print "now with proper uniform range"
delv = vmax-vmin
ad_result = skgof.ad_test(pits[mask], stats.uniform(loc=vmin,scale=delv))
return ad_result.statistic, ad_result.pvalue
def computeAD(data,mu,sd,seed):
np.random.seed(seed)
from skgof import ad_test
from scipy.stats import norm, anderson
res = ad_test(data, norm(loc=mu,scale=sd))
res2 = anderson(data, 'norm')
return [res.statistic, res.pvalue,res2.critical_values.tolist()]
def _p(test_i, null_i, M_i, d_i):
gpd_fit = None
gpd_fit_p_value = None
n_i = n
# TODO: no need to sort as much as N numbers, do partial sort:
# but this requires some tests (both performance and unit)
# null_i_partitioned = np.partition(null_i, n_i+1)
# null_i_first_n_sorted = sorted(null_i_partitioned[:-n_i+1])
null_i = sorted(null_i)
t = None
if all(np.isnan(null_i)):
return np.nan, False, np.nan, np.nan
# compute ecdf based, biased estimate of p-value
raw_ecdf_estimate = (ecdf_pseudocount + d_i.sum()) / (N + 1)
if M_i < m:
# fit GDP, reducing $n$ until convergance
while n_i > 0:
# -1 because Python has 0-based indexing
t = (null_i[-n_i-1] + null_i[-n_i-2]) / 2
y_untill_n = null_i[-n_i:]
exceedences = y_untill_n - t
assert all(y_untill_n >= t)
assert len(exceedences) == n_i
fit = genpareto.fit(exceedences)
fitted = genpareto(*fit)
gpd_fit = fitted
gpd_fit_p_value = ad_test(exceedences, fitted).pvalue
if gpd_fit_p_value <= 0.05:
break
else:
n_i -= decrease_n_by
if gpd_fit and gpd_fit_p_value < 0.05:
return n_i / N * (1 - gpd_fit.cdf(test_i - t)), True, gpd_fit_p_value, raw_ecdf_estimate
else:
if gpd_fit:
# TODO: get index and highlight which observation could not be fitted!
warn(f'A good GPD fit could not be reached, using ECDF estimate instead')
return raw_ecdf_estimate, False, np.nan, raw_ecdf_estimate
def NZAD(self, vmin=0.005, vmax=1.995):
"""
Compute the Anderson Darling statistic and p-value for the
two distributions of sumpz and true_z vector of spec-z's
Parameters:
vmin, vmax: specz values outside of these values are discarded
-----------
using: string
which parameterization to evaluate
Returns:
--------
Anderson-Darling statistic and pvalue
"""
#copy the form of Rongpu's use of skgof functions
#will have to use QPPDFCDF class, as those expect objects
#that have a .cdf method for a vector of values
print "using %f and %f for vmin and vmax\n" % (vmin, vmax)
szs = self.truth
mask = (szs > vmin) & (szs < vmax)
tmpnzfunc = QPPDFCDF(self.stackpz, self.dx)
nzAD = skgof.ad_test(szs[mask], tmpnzfunc)
return nzAD.statistic, nzAD.pvalue