def test_compute_grouped_elementary_effects():
model_inputs = np.array([[.39, -.39, -1.64, 0.39, -0.39, -0.39, 0.39, 0.39,
-1.64, -0.39, 0.39, -1.64, 1.64, 1.64, 1.64],
[-1.64, 1.64, 0.39, -1.64, 1.64, 1.64, -1.64,
-1.64, -1.64, 1.64, 0.39, -1.64, 1.64, 1.64,
1.64],
[-1.64, 1.64, 0.39, -1.64, 1.64, 1.64, -1.64,
-1.64, 0.39, 1.64, -1.64, 0.39, -.39, -.39, -.39]
])
model_results = np.array([13.85, -10.11, 1.12])
problem = {'names': ['1', '2', '3', '4', '5', '6', '7', '8',
'9', '10', '11', '12', '13', '14', '15'],
'bounds': [[]],
'groups':
(np.array([[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 0, 0, 0, 0, 0]]),
['gp1', 'gp2']),
'num_vars': 15
}
ee = compute_elementary_effects(model_inputs, model_results, 3, 2. / 3)
mu_star = np.average(np.abs(ee), axis=1)
actual = compute_grouped_metric(mu_star, problem['groups'][0].T)
desired = np.array([16.86, 35.95])
assert_allclose(actual, desired, atol=1e-1)
def test_compute_elementary_effects():
'''
Inputs for elementary effects taken from Exercise 5 from Saltelli (2008).
See page 140-145.
`model_inputs` are from trajectory t_1 from table 3.10 on page 141.
`desired` is equivalent to column t_1 in table 3.12 on page 145.
'''
model_inputs = np.array([
[1.64, -1.64, -1.64, 0.39, -0.39, 0.39, -1.64, -
1.64, -0.39, -0.39, 1.64, 1.64, -0.39, 0.39, 1.64],
[1.64, -1.64, -1.64, 0.39, -0.39, -1.64, -1.64, -
1.64, -0.39, -0.39, 1.64, 1.64, -0.39, 0.39, 1.64],
[1.64, -1.64, -1.64, 0.39, -0.39, -1.64, -1.64, -
1.64, 1.64, -0.39, 1.64, 1.64, -0.39, 0.39, 1.64],
[1.64, -1.64, -1.64, 0.39, -0.39, -1.64, -1.64,
0.39, 1.64, -0.39, 1.64, 1.64, -0.39, 0.39, 1.64],
[1.64, -1.64, -1.64, -1.64, -0.39, -1.64, -1.64,
0.39, 1.64, -0.39, 1.64, 1.64, -0.39, 0.39, 1.64],
[1.64, -1.64, -1.64, -1.64, -0.39, -1.64, -1.64,
0.39, 1.64, -0.39, 1.64, -0.39, -0.39, 0.39, 1.64],
[1.64, -1.64, -1.64, -1.64, -0.39, -1.64, -1.64,
0.39, 1.64, -0.39, 1.64, -0.39, -0.39, -1.64, 1.64],
[1.64, 0.39, -1.64, -1.64, -0.39, -1.64, -1.64,
0.39, 1.64, -0.39, 1.64, -0.39, -0.39, -1.64, 1.64],
[1.64, 0.39, -1.64, -1.64, -0.39, -1.64, -1.64, 0.39,
1.64, -0.39, 1.64, -0.39, -0.39, -1.64, -0.39],
[1.64, 0.39, -1.64, -1.64, -0.39, -1.64, -1.64,
0.39, 1.64, 1.64, 1.64, -0.39, -0.39, -1.64, -0.39],
[1.64, 0.39, -1.64, -1.64, 1.64, -1.64, -1.64, 0.39,
1.64, 1.64, 1.64, -0.39, -0.39, -1.64, -0.39],
[1.64, 0.39, -1.64, -1.64, 1.64, -1.64, -1.64, 0.39,
1.64, 1.64, -0.39, -0.39, -0.39, -1.64, -0.39],
[1.64, 0.39, -1.64, -1.64, 1.64, -1.64, 0.39, 0.39,
1.64, 1.64, -0.39, -0.39, -0.39, -1.64, -0.39],
[1.64, 0.39, 0.39, -1.64, 1.64, -1.64, 0.39, 0.39,
1.64, 1.64, -0.39, -0.39, -0.39, -1.64, -0.39],
[-0.39, 0.39, 0.39, -1.64, 1.64, -1.64, 0.39, 0.39,
1.64, 1.64, -0.39, -0.39, -0.39, -1.64, -0.39],
[-0.39, 0.39, 0.39, -1.64, 1.64, -1.64, 0.39, 0.39, 1.64, 1.64, -0.39, -0.39, 1.64, -1.64, -0.39]],
dtype=np.float)
model_outputs = np.array([24.9, 22.72, 21.04, 16.01, 10.4, 10.04, 8.6, 13.39, 4.69, 8.02, 9.98, 3.75, 1.33, 2.59, 6.37, 9.99],
dtype=np.float)
delta = 2. / 3
actual = compute_elementary_effects(model_inputs, model_outputs, 16, delta)
desired = np.array([[-5.67], [7.18], [1.89], [8.42], [2.93], [3.28], [-3.62], [-7.55],
[-2.51], [5.00], [9.34], [0.54], [5.43], [2.15], [13.05]],
dtype=np.float)
assert_allclose(actual, desired, atol=1e-1)
def test_compute_elementary_effects_small():
'''
Computes elementary effects for two variables,
over six trajectories with four levels.
'''
model_inputs = np.array([[0, 1. / 3], [0, 1], [2. / 3, 1],
[0, 1. / 3], [2. / 3, 1. / 3], [2. / 3, 1],
[2. / 3, 0], [2. / 3, 2. / 3], [0, 2. / 3],
[1. / 3, 1], [1, 1], [1, 1. / 3],
[1. / 3, 1], [1. / 3, 1. / 3], [1, 1. / 3],
[1. / 3, 2. / 3], [1. / 3, 0], [1, 0]],
dtype=np.float)
model_outputs = np.array([0.97, 0.71, 2.39, 0.97, 2.3, 2.39, 1.87, 2.40, 0.87, 2.15, 1.71, 1.54, 2.15, 2.17, 1.54, 2.2, 1.87, 1.0],
dtype=np.float)
delta = 2. / 3
actual = compute_elementary_effects(model_inputs, model_outputs, 3, delta)
desired = np.array(
[[2.52, 2.01, 2.30, -0.66, -0.93, -1.30], [-0.39, 0.13, 0.80, 0.25, -0.02, 0.51]])
assert_allclose(actual, desired, atol=1e-0)
def test_compute_elementary_effects_small():
'''
Computes elementary effects for two variables,
over six trajectories with four levels.
'''
model_inputs = np.array(
[[0, 1. / 3], [0, 1], [2. / 3, 1], [0, 1. / 3], [2. / 3, 1. / 3],
[2. / 3, 1], [2. / 3, 0], [2. / 3, 2. / 3], [0, 2. / 3], [1. / 3, 1],
[1, 1], [1, 1. / 3], [1. / 3, 1], [1. / 3, 1. / 3], [1, 1. / 3],
[1. / 3, 2. / 3], [1. / 3, 0], [1, 0]],
dtype=float)
model_outputs = np.array([
0.97, 0.71, 2.39, 0.97, 2.3, 2.39, 1.87, 2.40, 0.87, 2.15, 1.71, 1.54,
2.15, 2.17, 1.54, 2.2, 1.87, 1.0
],
dtype=float)
delta = 2. / 3
actual = compute_elementary_effects(model_inputs, model_outputs, 3, delta)
desired = np.array([[2.52, 2.01, 2.30, -0.66, -0.93, -1.30],
[-0.39, 0.13, 0.80, 0.25, -0.02, 0.51]])
assert_allclose(actual, desired, atol=1e-0)
def test_compute_elementary_effects():
'''
Inputs for elementary effects taken from Exercise 5 from Saltelli (2008).
See page 140-145.
`model_inputs` are from trajectory t_1 from table 3.10 on page 141.
`desired` is equivalent to column t_1 in table 3.12 on page 145.
'''
model_inputs = np.array(
[[
1.64, -1.64, -1.64, 0.39, -0.39, 0.39, -1.64, -1.64, -0.39, -0.39,
1.64, 1.64, -0.39, 0.39, 1.64
],
[
1.64, -1.64, -1.64, 0.39, -0.39, -1.64, -1.64, -1.64, -0.39,
-0.39, 1.64, 1.64, -0.39, 0.39, 1.64
],
[
1.64, -1.64, -1.64, 0.39, -0.39, -1.64, -1.64, -1.64, 1.64, -0.39,
1.64, 1.64, -0.39, 0.39, 1.64
],
[
1.64, -1.64, -1.64, 0.39, -0.39, -1.64, -1.64, 0.39, 1.64, -0.39,
1.64, 1.64, -0.39, 0.39, 1.64
],
[
1.64, -1.64, -1.64, -1.64, -0.39, -1.64, -1.64, 0.39, 1.64, -0.39,
1.64, 1.64, -0.39, 0.39, 1.64
],
[
1.64, -1.64, -1.64, -1.64, -0.39, -1.64, -1.64, 0.39, 1.64, -0.39,
1.64, -0.39, -0.39, 0.39, 1.64
],
[
1.64, -1.64, -1.64, -1.64, -0.39, -1.64, -1.64, 0.39, 1.64, -0.39,
1.64, -0.39, -0.39, -1.64, 1.64
],
[
1.64, 0.39, -1.64, -1.64, -0.39, -1.64, -1.64, 0.39, 1.64, -0.39,
1.64, -0.39, -0.39, -1.64, 1.64
],
[
1.64, 0.39, -1.64, -1.64, -0.39, -1.64, -1.64, 0.39, 1.64, -0.39,
1.64, -0.39, -0.39, -1.64, -0.39
],
[
1.64, 0.39, -1.64, -1.64, -0.39, -1.64, -1.64, 0.39, 1.64, 1.64,
1.64, -0.39, -0.39, -1.64, -0.39
],
[
1.64, 0.39, -1.64, -1.64, 1.64, -1.64, -1.64, 0.39, 1.64, 1.64,
1.64, -0.39, -0.39, -1.64, -0.39
],
[
1.64, 0.39, -1.64, -1.64, 1.64, -1.64, -1.64, 0.39, 1.64, 1.64,
-0.39, -0.39, -0.39, -1.64, -0.39
],
[
1.64, 0.39, -1.64, -1.64, 1.64, -1.64, 0.39, 0.39, 1.64, 1.64,
-0.39, -0.39, -0.39, -1.64, -0.39
],
[
1.64, 0.39, 0.39, -1.64, 1.64, -1.64, 0.39, 0.39, 1.64, 1.64,
-0.39, -0.39, -0.39, -1.64, -0.39
],
[
-0.39, 0.39, 0.39, -1.64, 1.64, -1.64, 0.39, 0.39, 1.64, 1.64,
-0.39, -0.39, -0.39, -1.64, -0.39
],
[
-0.39, 0.39, 0.39, -1.64, 1.64, -1.64, 0.39, 0.39, 1.64, 1.64,
-0.39, -0.39, 1.64, -1.64, -0.39
]],
dtype=float)
model_outputs = np.array([
24.9, 22.72, 21.04, 16.01, 10.4, 10.04, 8.6, 13.39, 4.69, 8.02, 9.98,
3.75, 1.33, 2.59, 6.37, 9.99
],
dtype=float)
delta = 2. / 3
actual = compute_elementary_effects(model_inputs, model_outputs, 16, delta)
desired = np.array(
[[-5.67], [7.18], [1.89], [8.42], [2.93], [3.28], [-3.62], [-7.55],
[-2.51], [5.00], [9.34], [0.54], [5.43], [2.15], [13.05]],
dtype=float)
assert_allclose(actual, desired, atol=1e-1)
problem = {
"num_vars" : len(par_morris),
"names" : par_morris,
"groups" : None,
"bounds" : [[0, lev-1]] * len(par_morris)
}
X = traj_id.iloc[idxs].values.astype(float)
#%%Get morris output
delta = lev / (2 * (lev - 1))
ee={}
for var in df.columns:
Y = df[var].values
ee[var] = compute_elementary_effects( #For bookkeeping purposes
X, Y, problem["num_vars"]+1, delta)
ee[var] = pd.DataFrame(data = ee[var].T, columns = problem["names"])
#%%Correct errors
#In one instance Kh_aqf was lower than Kh_aqt (trajectory 9),
#since we prescribed Kv_aqt and due to a high Kh_Kv, this got converted to a Kh_aqt higher than Kh_aqf.
#This flipper around the elementary effects for N_aqt and f_aqt, which we correct for here.
ee["old_water"]["N_aqt"][8] = ee["old_water"]["N_aqt"][8]*-1
ee["old_water"]["f_aqt"][8] = ee["old_water"]["f_aqt"][8]*-1
#
#ee["offshore_fw"]["alpha"][6] = 0
#%%Calculate Morris metrics
output = {}