def test_complex_model_complex_data():
# ======================================================================
"Check the fit of a complex-valued model to complex-valued data"
fitResult = fit(mymodel, y)
assert np.allclose(fitResult.model.real, y.real) and np.allclose(
fitResult.model.imag, y.imag)
def test_complex_model_real_data():
# ======================================================================
"Check the fit of a complex-valued model to real-valued data"
fitResult = fit(mymodel, y.real)
assert np.allclose(fitResult.model.real, y.real) and np.allclose(
fitResult.model.imag, np.zeros_like(y.real))
def test_ex_fwd5pdeer_fit():
"Check the forward 5-pulse DEER experimental model in fitting."
experiment = ex_fwd5pdeer(tau1,tau2,tau3, pathways=[1,2,3])
Vmodel = dipolarmodel(tdeer,r,Bmodel=bg_hom3d,experiment=experiment)
result = fit(Vmodel,Vfwd5pdeer,nonlin_tol=1e-3)
assert np.allclose(Vfwd5pdeer,result.model,atol=1e-2) and ovl(result.P/1e5,Pr)>0.975
def test_ex_4pdeer_fit():
"Check the 4-pulse DEER experimental model."
experiment = ex_4pdeer(tau1,tau2, pathways=[1,2])
Vmodel = dipolarmodel(tdeer,r,Bmodel=bg_hom3d,experiment=experiment)
result = fit(Vmodel,V4pdeer,nonlin_tol=1e-3)
assert np.allclose(V4pdeer,result.model,atol=1e-2) and ovl(result.P/1e5,Pr)>0.975
def test_fit_Pnonparametric():
"Check that the model with one dipolar pathway is correct"
Vmodel = dipolarmodel(t,r,Bmodel=bg_hom3d,npathways=1)
result = fit(Vmodel,V1path,nonlin_tol=1e-5)
assert ovl(result.P/1e5,Pr)>0.975
def test_ex_ridme_fit():
"Check the RIDME experimental model in fitting."
experiment = ex_ridme(tau1,tau2, pathways=[1,2,3])
Vmodel = dipolarmodel(tridme,r,Bmodel=bg_hom3d,experiment=experiment)
result = fit(Vmodel,Vridme,nonlin_tol=1e-3)
assert np.allclose(Vridme,result.model,atol=1e-2) and ovl(result.P/1e5,Pr)>0.975
def test_fit_1pathways():
"Check that the model can be correctly fitted with one dipolar pathway"
Vmodel = dipolarmodel(t,r,dd_gauss,bg_hom3d,npathways=1)
result = fit(Vmodel,V1path,nonlin_tol=1e-3)
assert np.allclose(result.model,V1path)
def test_preserve_original():
"Check that the original model is not changed by the function"
model = Model(gauss)
model.mean.par0 = 3
model.width.par0 = 0.2
_ = fit(model, mock_data)
assert model._parameter_list() == ['mean', 'width']
def test_fit_nonparametric():
#================================================================
"Check that a semiparametric model can be correctly fitted"
model = _getmodel('nonparametric')
fitResult = fit(model, mock_data)
assert np.allclose(fitResult.model, mock_data, atol=1e-3)
def test_CIs_nonparametric():
#================================================================
"Check the default confidence intervals of the fitted parameters"
model = _getmodel('semiparametric')
fitResult = fit(model, mock_data)
assert_attributes_cis(fitResult, ['amp1', 'amp2'])
def test_ex_sifter_fit():
"Check the SIFTER experimental model in fitting."
experiment = ex_sifter(tau1,tau2)
Vmodel = dipolarmodel(tsifter,r,Bmodel=bg_hom3d,experiment=experiment)
result = fit(Vmodel,Vsifter,nonlin_tol=1e-3)
assert np.allclose(Vsifter,result.model,atol=1e-2) and ovl(result.P/1e5,Pr)>0.975
def test_fit_weight_unbounded():
"Check fitting with a penalty with unbounded weight"
model = deepcopy(dd_gauss)
penaltyobj = Penalty(penalty_fcn,'icc')
result = fit(model,mock_data,x,penalties=penaltyobj)
assert not ovl(result.model,mock_data)>0.975
def test_fit_semiparametric_vec_constant():
#================================================================
"Check that a vectorized semiparametric model can be correctly fitted while specifying an axis"
model = _getmodel_axis('semiparametric_vec')
x = np.linspace(0, 10, 100)
fitResult = fit(model, mock_data_fcn(x), x, noiselvl=1e-10)
assert np.allclose(fitResult.model, mock_data_fcn(x), atol=1e-3)
def test_fit_semiparametric_constant():
#================================================================
"Check that a semiparametric model can be correctly fitted while specifying an axis"
model = _getmodel_axis('semiparametric')
x = np.linspace(0, 10, 200)
fitResult = fit(model, mock_data_fcn(x), x)
assert np.allclose(fitResult.model, mock_data_fcn(x))
def test_fit_nonparametric_frozen():
#================================================================
"Check that a semiparametric model can be correctly fitted"
model = _getmodel('nonparametric')
model.amp1.freeze(0.5)
fitResult = fit(model, mock_data)
assert np.allclose(fitResult.model, mock_data)
def test_bootCIs_nonparametric():
#================================================================
"Check the bootstrapped confidence intervals of the fitted parameters"
model = _getmodel('semiparametric')
noisydata = mock_data + whitegaussnoise(0.01, seed=1)
fitResult = fit(model, noisydata, bootstrap=3)
assert_attributes_cis(fitResult, ['amp1', 'amp2'])
def test_real_model_complex_data():
# ======================================================================
"Check the fit of a real-valued model to complex-valued data"
mymodel = Model(gauss)
mymodel.mean.set(par0=4, lb=1, ub=6)
mymodel.width.set(par0=0.2, lb=0.05, ub=5)
fitResult = fit(mymodel, y.real + 1j * np.zeros_like(y.imag))
assert np.allclose(fitResult.model.real, y.real) and np.allclose(
fitResult.model.imag, np.zeros_like(y.real))
def test_fit_weight_frozen():
"Check fitting with a penalty with frozen weight"
model = deepcopy(dd_gauss)
penaltyobj = Penalty(penalty_fcn,'icc')
penaltyobj.weight.freeze(0.00001)
result = fit(model,mock_data,x,penalties=penaltyobj)
assert ovl(result.model,mock_data)>0.975
def test_fit_icc():
"Check fitting with a penalty with ICC-selected weight"
model = deepcopy(dd_gauss)
penaltyobj = Penalty(penalty_fcn,'icc')
penaltyobj.weight.set(lb=1e-6,ub=1e1)
result = fit(model,mock_data,x,penalties=penaltyobj)
assert ovl(result.model,mock_data)>0.975
def test_fit_parametric_frozen():
#================================================================
"Check that a parametric model can be correctly fitted"
model = _getmodel('parametric')
model.mean1.freeze(3)
fitResult = fit(model, mock_data)
assert np.allclose(fitResult.model, mock_data)
def test_fit_weight_bounded():
"Check fitting with a penalty with bounded weight"
model = deepcopy(dd_gauss)
penaltyobj = Penalty(penalty_fcn,'aicc')
penaltyobj.weight.set(lb=1e-10,ub=1e1)
result = fit(model,mock_data,x,penalties=penaltyobj)
assert ovl(result.model,mock_data)>0.975
def test_fit_2pathways():
"Check that the model can be correctly fitted with two dipolar pathways"
Vmodel = dipolarmodel(t,r,dd_gauss,bg_hom3d,npathways=2)
Vmodel.reftime1.freeze(0)
Vmodel.reftime2.freeze(2)
result = fit(Vmodel,V2path,nonlin_tol=1e-3)
assert np.allclose(result.model,V2path)
def test_fit_fullyfrozen_linear():
#================================================================
"Check that a model with all linear parameters frozen can be fitted"
model = _getmodel('semiparametric')
model.amp1.freeze(0.5)
model.amp2.freeze(0.6)
fitResult = fit(model, mock_data)
assert np.allclose(fitResult.model, mock_data)
def test_relate_fit():
"Check that the method works correctly for one related parameter"
model = dl.dd_gauss
x = np.linspace(0,10,400)
ref = model(x,4,0.2)
newmodel = relate(model, width = lambda mean: mean/20)
result = fit(newmodel,ref,x)
assert np.allclose(result.model,ref)
def test_fit_model():
"Check that that merge works correctly for two models"
model1 = dl.dd_gauss
model2 = dl.dd_rice
model = merge(model1, model2)
x = np.linspace(0, 10, 400)
truth = [model1(x, 3, 0.2), model2(x, 4, 0.5)]
result = fit(model, truth, x, x, weights=[1, 1])
assert np.allclose(result.model[0], truth[0]) and np.allclose(
result.model[1], truth[1])
def test_fit_propagate_semiparametric_vec():
#================================================================
"Check the propagate method of the fitResult object for evaluation of a vectorized semiparametric model"
model = _getmodel_axis('semiparametric_vec')
x = np.linspace(0, 10, 200)
fitResult = fit(model, mock_data_fcn(x), x)
modeluq = fitResult.propagate(model, x, lb=np.zeros_like(x))
assert_cis(modeluq)
def test_fit_evaluate_semiparametric_vec():
#================================================================
"Check the evaluate method of the fitResult object for evaluation of a vectorized semiparametric model"
model = _getmodel_axis('semiparametric_vec')
x = np.linspace(0, 10, 200)
fitResult = fit(model, mock_data_fcn(x), x)
response = fitResult.evaluate(model, x)
assert np.allclose(response, mock_data_fcn(x))
def test_fit_propagate_nonparametric_vec_bootstrapped():
#================================================================
"Check the propagate method of the fitResult object for evaluation of a vectorized nonparametric model"
model = _getmodel_axis('nonparametric_vec', vec=80)
x = np.linspace(0, 10, 80)
fitResult = fit(model, mock_data_fcn(x), x, bootstrap=3)
modeluq = fitResult.propagate(model, x, lb=np.zeros_like(x))
assert_cis(modeluq)
def test_fit_multiple_penalties():
"Check fitting with multiple penalties"
model = deepcopy(dd_gauss)
penaltyobj = Penalty(penalty_fcn,'icc')
penaltyobj2 = deepcopy(penaltyobj)
penaltyobj.weight.freeze(0.00001)
penaltyobj2.weight.freeze(0.00001)
result = fit(model,mock_data,x,penalties=[penaltyobj,penaltyobj2])
assert ovl(result.model,mock_data)>0.975
def test_fit_fullyfrozen_nonlinear():
#================================================================
"Check that a model with all nonlinear parameters frozen can be fitted"
model = _getmodel('semiparametric')
model.mean1.freeze(3)
model.mean2.freeze(4)
model.width1.freeze(0.5)
model.width2.freeze(0.2)
fitResult = fit(model, mock_data)
assert np.allclose(fitResult.model, mock_data)
