def test_insert_theta():
tau_in = [0.1, 1.0]
epdf = exponentials.ExponentialPDF(tau=tau_in)
theta_in = [0.2, 2., 0.3]
epdf.theta = theta_in
print(epdf.theta)
print(epdf.pars)
pars_in = [0.2, 2., 0.3, 0.7]
npt.assert_almost_equal(epdf.pars, pars_in)
def test_fixed_pars():
tau_in = [0.1, 1.0]
epdf = exponentials.ExponentialPDF(tau=tau_in)
epdf.fixed[1] = True
fixed = [False, True, False, True]
assert fixed == epdf.fixed
theta_in = [0.2, 0.3]
epdf.theta = theta_in
print(epdf.theta)
print(epdf.pars)
pars_in = [0.2, 1., 0.3, 0.7]
npt.assert_almost_equal(epdf.pars, pars_in)
npt.assert_almost_equal(epdf.theta, theta_in)
def test_area_none():
tau_in = [0.1, 1.0]
epdf = exponentials.ExponentialPDF(tau=tau_in)
area = np.ones(2) / 2
print(epdf.area)
npt.assert_almost_equal(epdf.area, area)
pars = np.concatenate((np.asarray(tau_in), area))
print(epdf.pars)
npt.assert_almost_equal(epdf.pars, pars)
print(epdf.fixed)
fixed = [False, False, False, True]
assert fixed == epdf.fixed
print(epdf.theta)
theta = pars[:-1]
npt.assert_almost_equal(epdf.theta, theta)
def histogram_xlog_ysqrt_data(X,
tres,
pdf=None,
tcrit=None,
xlabel='Dwell times'):
""" Plot dwell time histogram in log x and square root y. """
xout, yout = prepare_xlog_hist(X, tres)
fig = plt.figure(figsize=(3, 3))
ax = fig.add_subplot(111)
ax.semilogx(xout, np.sqrt(yout))
ax.set_xlabel(xlabel)
ax.set_ylabel('sqrt(frequency)')
t = np.logspace(math.log10(tres), math.log10(2 * max(X)), 512)
if pdf is None:
pdf = exponentials.ExponentialPDF([np.mean(X)], [1.0])
scale = __exponential_scale_factor(X, pdf, tres)
ax.plot(t, np.sqrt(scale * t * pdf.exp(pdf.theta, t)), '-b')
for ta, ar in zip(pdf.tau, pdf.area):
ax.plot(t, np.sqrt(scale * t * (ar / ta) * np.exp(-t / ta)), '--b')
if tcrit is not None:
for tc in np.asarray(tcrit):
ax.axvline(x=tc, color='g')
def setUp(self):
self.infile = "./tests/intervals.txt"
self.intervals = np.loadtxt(self.infile)
self.tau, self.area = [0.036, 1.1], [0.20]
self.theta = self.tau + self.area
self.epdf = exponentials.ExponentialPDF(self.tau, self.area)
self.res = minimize(self.epdf.LL,
self.epdf.theta,
args=self.intervals,
method='Nelder-Mead')
self.asd = errors.ApproximateSD(self.res.x, self.epdf.LL,
self.intervals)
#self.hess = asd.hess
#self.hess = eklib.hessian(self.res.x, self.epdf.LL, self.intervals)
#self.cov = eklib.covariance_matrix(self.theta, eklib.LLexp, self.intervals)
#self.cov = asd.covariance
#self.cov = nplin.inv(self.hess)
#self.appSD = asd.sd #np.sqrt(self.cov.diagonal())
#self.corr = asd.correlations #eklib.correlation_matrix(self.cov)
m = 2.0 # corresponds roughly to 2 SD
self.likints = errors.LikelihoodIntervals(self.res.x, self.epdf,
self.intervals, self.asd.sd,
m)