def nonlinear_fit_with_independent_as_t():
"""Test of PHS3000 nonlinear fit using an independent variable that was not 'x'"""
### Get results ###
import monashspa.PHS3000 as spa
# load the data
data = spa.tutorials.fitting.part_b_data
# slice the data into columns
t = data[:,0]
A = data[:,1]
u_A = data[:,2]
model = spa.make_lmfit_model("A_0*exp(-l*t)", independent_vars=["t"])
params = model.make_params(A_0=30, l=.005)
params.add('halflife', expr="log(2)/l")
fit_results = spa.model_fit(model, params, x=t, y=A, u_y=u_A)
results = spa.get_fit_parameters(fit_results)
### Expected results ###
expected_results = {
'A_0': 20.573035990441486,
'u_A_0': 0.6181473325960677,
'l': 0.005004779941925933,
'u_l': 0.00015840653659960648,
'halflife': 138.49703455557113,
'u_halflife': 4.38357646646529,
}
precision = 2e-7
### Check results match within precision ###
success = compare_dictionary(results, expected_results, precision)
return success
def linear_fit():
"""Basic test of PHS3000 linear fit"""
### Get results ###
import monashspa.PHS3000 as spa
# load the data
data = spa.tutorials.fitting.part_b_data
# slice the data into columns
t = data[:,0]
A = data[:,1]
u_A = data[:,2]
fit_results = spa.linear_fit(t, np.log(A), u_y=u_A/A)
results = spa.get_fit_parameters(fit_results)
### Expected results ###
expected_results = {
'intercept': 3.02843895796932,
'u_intercept': 0.030620495054301592,
'slope': -0.004897938859868103,
'u_slope': 0.0001633020007207088,
}
precision = 1e-7
### Check results match within precision ###
success = compare_dictionary(results, expected_results, precision)
return success
def nonlinear_fit_part_c2():
"""Test of PHS3000 non-linear fit to multi-model nuclear decay data"""
import monashspa.PHS3000 as spa
# load the data
data = spa.tutorials.fitting.part_c2_data
# slice the data into columns
t = data[:,0]
A = data[:,1]
u_A = data[:,2]
ag110_model = spa.make_lmfit_model("A_0*exp(-l*x)", prefix="AG110_")
ag108_model = spa.make_lmfit_model("A_0*exp(-l*x)", prefix="AG108_")
offset = spa.make_lmfit_model("c+x*0", name="offset")
model = ag110_model + ag108_model + offset
params = model.make_params(AG110_A_0=A[0], AG110_l=np.log(2)/t[4], AG108_l=np.log(2)/t[9], c=0)
params.add('c', min=0, value=0, max=0.1, vary=True)
params.add('AG110_halflife', expr='log(2)/AG110_l')
params.add('AG108_halflife', expr='log(2)/AG108_l')
fit_results = spa.model_fit(model, params, x=t, y=A, u_y=u_A)
results = spa.get_fit_parameters(fit_results)
expected_results = {
'AG110_A_0': 106.62466958768731,
'u_AG110_A_0': 3.7021128718124436,
'AG110_l': 0.03272908958050027,
'u_AG110_l': 0.0019419683658320412,
'AG108_A_0': 24.504027101025645,
'u_AG108_A_0': 2.492813131791383,
'AG108_l': 0.005097263436001847,
'u_AG108_l': 0.000586527763335435,
'c': 6.436383415431291e-06,
'u_c': 1.1676201,
'AG110_halflife': 21.178321470112536,
'u_AG110_halflife': 1.2566078265666092,
'AG108_halflife': 135.98417842489044,
'u_AG108_halflife': 15.647316857620634
}
precision = 5e-3
# split off the check for u_c as the precision (variation across platforms) is much worse than everything else.
# This is to be expected with multi-model non-linear fits.
u_c_precision = 3.0
u_c_result = {'u_c':results['u_c']}
u_c_expected_result = {'u_c':expected_results['u_c']}
del results['u_c']
del expected_results['u_c']
### Check results match within precision ###
success = compare_dictionary(results, expected_results, precision) and compare_dictionary(u_c_result, u_c_expected_result, u_c_precision)
return success
def non_linear_fit_part_c():
"""Test of PHS3000 non-linear fit to generic multi-model data"""
### Get results ###
import monashspa.PHS3000 as spa
# load the data
data = spa.tutorials.fitting.part_c1_data
# slice the data into columns
x = data[:,0]
y = data[:,1]
# Define the models
model1 = spa.make_lmfit_model("A1*exp(-(x-B1)**2/D1**2)", name="Gaussian1")
model2 = spa.make_lmfit_model("A2*exp(-(x-B2)**2/D2**2)", name="Gaussian2")
model3 = spa.make_lmfit_model("c+x*0", name="Offset")
model = model1 + model2 + model3
params = model.make_params(A1=np.max(y), B1=100, D1=6, A2=np.max(y), B2=160, c=3)
params.add('D2', expr='D1')
params.add('FWHM', expr='2*sqrt(log(2)*D1)')
fit_results = spa.model_fit(model, params, x=x, y=y)
results = spa.get_fit_parameters(fit_results)
### Expected results ###
expected_results = {
'A1': 16.58279101985785,
'u_A1': 0.351003396364092,
'B1': 95.82492417888295,
'u_B1': 0.12363799249150613,
'D1': 6.705149651134903,
'u_D1': 0.1412232821933916,
'A2': 13.690661831671,
'u_A2': 0.33945028545509237,
'B2': 164.19308257427863,
'u_B2': 0.1497567729486557,
'D2': 6.705149651134903,
'u_D2': 0.14122328218157232,
'c': 2.636117481366362,
'u_c': 0.06764023986252798,
'FWHM': 4.311684392863958,
'u_FWHM': 0.045406161696634036
}
precision = 1e-5
### Check results match within precision ###
success = compare_dictionary(results, expected_results, precision)
return success
def linear_fit_dual_custom_named_model():
"""Test of PHS3000 linear fit using the sum of two custom models with prefixes"""
### Get results ###
import monashspa.PHS3000 as spa
# load the data
data = spa.tutorials.fitting.part_b_data
# slice the data into columns
t = data[:,0]
A = data[:,1]
u_A = data[:,2]
model1 = spa.make_lmfit_model("m*x", prefix="m1")
model2 = spa.make_lmfit_model("c+x*0", prefix="m2")
model = model1 + model2
params = model.make_params(m1m=0.04, m2c=5)
params.add('halflife', expr="-log(2)/m1m")
params.add('A_0', expr="exp(m2c)")
fit_results = spa.model_fit(model, params, x=t, y=np.log(A), u_y=u_A/A)
results = spa.get_fit_parameters(fit_results)
### Expected results ###
expected_results = {
'm2c': 3.02843895796932,
'u_m2c': 0.030620495054301592,
'm1m': -0.004897938859868103,
'u_m1m': 0.0001633020007207088,
'halflife': 141.51813658584769,
'u_halflife': 4.718351025589936,
'A_0': 20.664948544330286,
'u_A_0': 0.6327709546990624,
}
precision = 1e-5
### Check results match within precision ###
success = compare_dictionary(results, expected_results, precision)
return success