def test_fit_minimal():
output = mc3.fit(data, uncert, quad, np.copy(params), indparams=[x])
np.testing.assert_allclose(output['best_log_post'], -54.43381306220858)
np.testing.assert_equal(-2 * output['best_log_post'], output['best_chisq'])
np.testing.assert_allclose(output['bestp'],
np.array([4.28263253, -2.40781859, 0.49534411]),
rtol=1e-7)
def test_fit_trf():
output = mc3.fit(data,
uncert,
quad,
np.copy(params),
indparams=[x],
leastsq='trf')
np.testing.assert_allclose(output['best_log_post'], -54.43381306220856)
np.testing.assert_allclose(output['bestp'],
np.array([4.28263252, -2.40781858, 0.49534411]),
rtol=1e-7)
def test_fit_leastsq_error(capsys):
with pytest.raises(SystemExit):
output = mc3.fit(data,
uncert,
quad,
np.copy(params),
indparams=[x],
leastsq='invalid')
captured = capsys.readouterr()
assert "Invalid 'leastsq' input (invalid). Must select from " \
"['lm', 'trf']." in captured.out
def test_fit_bounds():
output = mc3.fit(data,
uncert,
quad, [4.5, -2.5, 0.5],
indparams=[x],
pmin=[4.4, -3.0, 0.4],
pmax=[5.0, -2.0, 0.6],
leastsq='trf')
np.testing.assert_allclose(output['best_log_post'], -54.45536109795812)
np.testing.assert_allclose(output['bestp'],
np.array([4.4, -2.46545897, 0.5009366]),
rtol=1e-7)
def test_fit_shared():
output = mc3.fit(data1,
uncert1,
quad,
np.copy(params),
indparams=[x],
pstep=[1.0, -1, 1.0])
assert output['bestp'][1] == output['bestp'][0]
np.testing.assert_allclose(output['best_log_post'], -51.037667264657)
np.testing.assert_allclose(output['bestp'],
np.array([4.58657213, 4.58657213, 0.43347714]),
rtol=1e-7)
def test_fit_fixed():
pars = np.copy(params)
pars[0] = p0[0]
output = mc3.fit(data,
uncert,
quad,
pars,
indparams=[x],
pstep=[0.0, 1.0, 1.0])
assert output['bestp'][0] == pars[0]
np.testing.assert_allclose(output['best_log_post'], -54.507722717665466)
np.testing.assert_allclose(output['bestp'],
np.array([4.5, -2.51456999, 0.50570154]),
rtol=1e-7)
def test_fit_priors():
prior = np.array([4.5, 0.0, 0.0])
priorlow = np.array([0.1, 0.0, 0.0])
priorup = np.array([0.1, 0.0, 0.0])
output = mc3.fit(data,
uncert,
quad,
np.copy(params),
indparams=[x],
prior=prior,
priorlow=priorlow,
priorup=priorup)
np.testing.assert_allclose(output['best_log_post'], -54.50548056991611)
# First parameter is closer to 4.5 than without a prior:
np.testing.assert_allclose(output['bestp'],
np.array([4.49340587, -2.51133157, 0.50538734]),
rtol=1e-7)
names = [None] * ntargets
flat_chisq = np.zeros(ntargets)
flat_rchisq = np.zeros(ntargets)
flat_bic = np.zeros(ntargets)
for i in range(ntargets):
cfile = [f for f in os.listdir(targets[i]) if f.endswith(".cfg")][0]
with cd(targets[i]):
pyrat = pb.run(cfile, run_step='dry', no_logfile=True)
data = pyrat.obs.data
uncert = pyrat.obs.uncert
indparams = [data]
params = np.array([np.mean(data)])
pstep = np.array([1.0])
mc3_fit = mc3.fit(data, uncert, sfit, params, indparams, pstep)
dof = len(data) - len(params)
names[i] = cfile.split('_')[1] + ' ' + cfile.split('_')[2]
flat_chisq[i] = mc3_fit['best_chisq']
flat_rchisq[i] = mc3_fit['best_chisq'] / dof
flat_bic[i] = mc3_fit['best_chisq'] + len(params) * np.log(len(data))
with open("stats/flat_fit.dat", "w") as f:
f.write("Flat-curve fit to transmission data (Nfree = 1):\n\n"
"Planet chi-square red-chisq BIC\n"
"----------------------------------------------------\n")
for i in range(ntargets):
f.write(f"{names[i]:20s} {flat_chisq[i]:10.3f} "
f"{flat_rchisq[i]:9.3f} {flat_bic[i]:7.3f}\n")