def test_conditional_loss_duplicates(self):
# we feed compute_conditional_loss with some duplicated data to see if
# it's handled correctly
loss1 = scientific.conditional_loss(Curve([
(0.21, 0.131), (0.24, 0.108),
(0.27, 0.089), (0.30, 0.066),
]), 0.100)
# duplicated y values, different x values, (0.19, 0.131), (0.20, 0.131)
# should be skipped
loss2 = scientific.conditional_loss(Curve([
(0.19, 0.131), (0.20, 0.131), (0.21, 0.131),
(0.24, 0.108), (0.27, 0.089), (0.30, 0.066),
]), 0.100)
numpy.testing.assert_allclose(loss1, loss2)
def test_conditional_loss_computation(self):
loss_curve = Curve([
(0.21, 0.131), (0.24, 0.108),
(0.27, 0.089), (0.30, 0.066),
])
self.assertAlmostEqual(0.2526, scientific.conditional_loss(
loss_curve, 0.100), 4)
def test_loss_is_max_if_probability_is_too_low(self):
loss_curve = Curve([
(0.21, 0.131), (0.24, 0.108),
(0.27, 0.089), (0.30, 0.066),
])
self.assertAlmostEqual(0.30,
scientific.conditional_loss(loss_curve, 0.050))
def __call__(self, asset, hazard):
asset_output = self.loss_curve_calculator(asset, hazard)
cl = dict((poe,
scientific.conditional_loss(asset_output.loss_curve, poe))
for poe in self.conditional_loss_poes)
return asset_output._replace(conditional_losses=cl)
