def get_weighted_poes(gsim,
sctx,
rctx,
dctx,
imt,
imls,
truncation_level,
weighting=DEFAULT_WEIGHTING):
"""
This function implements the NGA West 2 GMPE epistemic uncertainty
adjustment factor without re-calculating the actual GMPE each time.
:param gsim:
Instance of the GMPE
:param list weighting:
Weightings as a list of tuples of (weight, number standard deviations
of the epistemic uncertainty adjustment)
"""
if truncation_level is not None and truncation_level < 0:
raise ValueError('truncation level must be zero, positive number '
'or None')
gsim._check_imt(imt)
adjustment = nga_west2_epistemic_adjustment(rctx.mag, dctx.rrup)
adjustment = adjustment.reshape(adjustment.shape + (1, ))
if truncation_level == 0:
# zero truncation mode, just compare imls to mean
imls = gsim.to_distribution_values(imls)
mean, _ = gsim.get_mean_and_stddevs(sctx, rctx, dctx, imt, [])
mean = mean.reshape(mean.shape + (1, ))
output = np.zeros([mean.shape[0], imls.shape[0]])
for (wgt, fct) in weighting:
output += (wgt * (imls <=
(mean + (fct * adjustment))).astype(float))
return output
else:
# use real normal distribution
assert (const.StdDev.TOTAL
in gsim.DEFINED_FOR_STANDARD_DEVIATION_TYPES)
imls = gsim.to_distribution_values(imls)
mean, [stddev] = gsim.get_mean_and_stddevs(sctx, rctx, dctx, imt,
[const.StdDev.TOTAL])
mean = mean.reshape(mean.shape + (1, ))
stddev = stddev.reshape(stddev.shape + (1, ))
output = np.zeros([mean.shape[0], imls.shape[0]])
for (wgt, fct) in weighting:
values = (imls - (mean + (fct * adjustment))) / stddev
if truncation_level is None:
output += (wgt * _norm_sf(values))
else:
output += (wgt * _truncnorm_sf(truncation_level, values))
return output
def get_weighted_poes(gsim, sctx, rctx, dctx, imt, imls, truncation_level,
weighting=DEFAULT_WEIGHTING):
"""
This function implements the NGA West 2 GMPE epistemic uncertainty
adjustment factor without re-calculating the actual GMPE each time.
:param gsim:
Instance of the GMPE
:param list weighting:
Weightings as a list of tuples of (weight, number standard deviations
of the epistemic uncertainty adjustment)
"""
if truncation_level is not None and truncation_level < 0:
raise ValueError('truncation level must be zero, positive number '
'or None')
gsim._check_imt(imt)
adjustment = nga_west2_epistemic_adjustment(rctx.mag, dctx.rrup)
adjustment = adjustment.reshape(adjustment.shape + (1, ))
if truncation_level == 0:
# zero truncation mode, just compare imls to mean
imls = gsim.to_distribution_values(imls)
mean, _ = gsim.get_mean_and_stddevs(sctx, rctx, dctx, imt, [])
mean = mean.reshape(mean.shape + (1, ))
output = np.zeros([mean.shape[0], imls.shape[0]])
for (wgt, fct) in weighting:
output += (wgt *
(imls <= (mean + (fct * adjustment))).astype(float))
return output
else:
# use real normal distribution
assert (const.StdDev.TOTAL
in gsim.DEFINED_FOR_STANDARD_DEVIATION_TYPES)
imls = gsim.to_distribution_values(imls)
mean, [stddev] = gsim.get_mean_and_stddevs(sctx, rctx, dctx, imt,
[const.StdDev.TOTAL])
mean = mean.reshape(mean.shape + (1, ))
stddev = stddev.reshape(stddev.shape + (1, ))
output = np.zeros([mean.shape[0], imls.shape[0]])
for (wgt, fct) in weighting:
values = (imls - (mean + (fct * adjustment))) / stddev
if truncation_level is None:
output += (wgt * _norm_sf(values))
else:
output += (wgt * _truncnorm_sf(truncation_level, values))
return output