def test_estimate(self):
print "Testing parameter estimation of GammaP distribution ..."
sys.stdout.flush()
myu = 10 * np.random.rand(1)[0]
mys = 10 * np.random.rand(1)[0]
myp = 1.5 * np.random.rand() + .5
p = Distributions.GammaP({'u': myu, 's': mys, 'p': myp})
dat = p.sample(10000)
p = Distributions.GammaP()
p.estimate(dat)
self.assertFalse(np.abs(p.param['u'] - myu) > self.TolParam,\
'Difference in Shape parameter for Gamma distribution greater than ' + str(self.TolParam['u']))
self.assertFalse(np.abs(p.param['s'] - mys) > self.TolParam,\
'Difference in Scale parameter for Gamma distribution greater than ' + str(self.TolParam['s']))
self.assertFalse(np.abs(p.param['p'] - mys) > self.TolParam,\
'Difference in Scale parameter for Gamma distribution greater than ' + str(self.TolParam['p']))
def test_derivatives(self):
print "Testing derivatives w.r.t. data ... "
sys.stdout.flush()
myu = 10 * np.random.rand(1)[0]
mys = 10 * np.random.rand(1)[0]
myp = 1.5 * np.random.rand() + .5
p = Distributions.GammaP({'u': myu, 's': mys, 'p': myp})
dat = p.sample(100)
h = 1e-8
tol = 1e-4
y = np.array(dat.X) + h
df = p.dldx(dat)
df2 = (p.loglik(Data(y)) - p.loglik(dat)) / h
self.assertFalse(
np.max(np.abs(df - df2)) > tol,
'Difference in derivative of log-likelihood for GammaP greater than '
+ str(tol))