def setUp(self):
data = south()
data['n_samples'] = 0
np.random.seed(TEST_SEED)
self.trace = Trace.from_csv(FULL_PATH + '/data/south_mvcm_5000', multi=True)
self.single_trace = Trace.from_csv(FULL_PATH + '/data/south_mvcm_5000_0.csv')
self.size_known = json.load(open(FULL_PATH + '/data/effective_size.json', 'r'))
def setUp(self):
self.a = {chr(i+97):list(range(10)) for i in range(5)}
self.t = Trace(**self.a)
self.mt = Trace(self.a,self.a,self.a)
self.real_mt = Trace.from_csv(FULL_PATH + r'/data/south_mvcm_5000', multi=True)
self.real_singles = [Trace.from_csv(FULL_PATH + r'/data/south_mvcm_5000_{}.csv'
.format(i)) for i in range(4)]
def setUp(self):
data = south()
data['n_samples'] = 0
np.random.seed(TEST_SEED)
self.trace = Trace.from_csv(FULL_PATH + '/data/south_mvcm_5000',
multi=True)
self.single_trace = Trace.from_csv(FULL_PATH +
'/data/south_mvcm_5000_0.csv')
self.geweke_known = json.load(open(FULL_PATH + '/data/geweke.json'))
def setUp(self):
data = south()
data['n_samples'] = 0
np.random.seed(TEST_SEED)
test_methods = ['obm', 'bm', 'bartlett', 'hanning', 'tukey']
self.trace = Trace.from_csv(FULL_PATH + '/data/south_mvcm_5000', multi=True)
self.single_trace = Trace.from_csv(FULL_PATH + '/data/south_mvcm_5000_0.csv')
self.bm = json.load(open(FULL_PATH + '/data/mcse_bm.json', 'r'))
self.obm = json.load(open(FULL_PATH + '/data/mcse_obm.json', 'r'))
self.tukey = json.load(open(FULL_PATH + '/data/mcse_hanning.json', 'r'))
self.bartlett = json.load(open(FULL_PATH + '/data/mcse_bartlett.json', 'r'))
self.hanning = self.tukey
def setUp(self):
super(Test_Lower_SMA, self).build_self()
self.cls = lower.Lower_SMA
del self.inputs["M"]
self.inputs['n_samples'] = 0
instance = self.cls(**self.inputs)
self.answer_trace = Trace.from_csv(FULL_PATH + '/data/lower_sma.csv')
def setUp(self):
super(Test_Lower_SE, self).build_self()
self.cls = lower.Lower_SE
del self.inputs["M"]
self.inputs['n_samples'] = 0
instance = self.cls(**self.inputs)
self.answer_trace = Trace.from_csv(FULL_PATH + '/data/lower_se.csv')
def setUp(self):
super(Test_Upper_SE, self).build_self()
self.cls = upper.Upper_SE
del self.inputs["W"]
self.inputs['n_samples'] = 0
instance = self.cls(**self.inputs)
self.answer_trace = Trace.from_csv(FULL_PATH + '/data/upper_se.csv')
def setUp(self):
super(Test_Upper_SMA, self).build_self()
self.cls = upper.Upper_SMA
del self.inputs["W"]
self.inputs['n_samples'] = 0
instance = self.cls(**self.inputs)
self.answer_trace = Trace.from_csv(FULL_PATH + '/data/upper_sma.csv')
def setUp(self):
data = south()
data['n_samples'] = 0
np.random.seed(TEST_SEED)
test_methods = ['obm', 'bm', 'bartlett', 'hanning', 'tukey']
self.trace = Trace.from_csv(FULL_PATH + '/data/south_mvcm_5000',
multi=True)
self.single_trace = Trace.from_csv(FULL_PATH +
'/data/south_mvcm_5000_0.csv')
self.bm = json.load(open(FULL_PATH + '/data/mcse_bm.json', 'r'))
self.obm = json.load(open(FULL_PATH + '/data/mcse_obm.json', 'r'))
self.tukey = json.load(open(FULL_PATH + '/data/mcse_hanning.json',
'r'))
self.bartlett = json.load(
open(FULL_PATH + '/data/mcse_bartlett.json', 'r'))
self.hanning = self.tukey
def setUp(self):
super(Test_MVCM, self).build_self()
self.cls = M.MVCM
del self.inputs['M']
del self.inputs['W']
self.inputs['n_samples'] = 0
self.instance = self.cls(**self.inputs)
self.answer_trace = Trace.from_csv(FULL_PATH + '/data/mvcm.csv')
def setUp(self):
super(Test_MVCM, self).build_self()
self.cls = M.MVCM
del self.inputs['M']
del self.inputs['W']
self.inputs['n_samples'] = 0
self.instance = self.cls(**self.inputs)
self.answer_trace = Trace.from_csv(FULL_PATH + '/data/mvcm.csv')
def test_mvcm(self):
instance = self.cls(**self.inputs)
np.random.seed(TEST_SEED)
instance.draw()
other_answers = Trace.from_csv(FULL_PATH + '/data/mvcm.csv')
strip_out = [col for col in instance.trace.varnames if col not in other_answers.varnames]
other_answers._assert_allclose(instance.trace.drop(
*strip_out, inplace=False),
rtol=RTOL, atol=ATOL)
def setUp(self):
data = south()
data['n_samples'] = 0
with open(FULL_PATH + '/data/psrf_noburn.json', 'r') as noburn:
self.noburn = json.load(noburn)
with open(FULL_PATH + '/data/psrf_brooks.json', 'r') as brooks:
self.known_brooks = json.load(brooks)
with open(FULL_PATH + '/data/psrf_gr.json', 'r') as gr:
self.known_gr = json.load(gr)
np.random.seed(TEST_SEED)
self.trace = Trace.from_csv(FULL_PATH + '/data/south_mvcm_5000', multi=True)
self.mockmodel = Hashmap(trace=self.trace)
def test_mvcm(self):
instance = self.cls(**self.inputs)
np.random.seed(TEST_SEED)
instance.draw()
other_answers = Trace.from_csv(FULL_PATH + '/data/mvcm.csv')
strip_out = [
col for col in instance.trace.varnames
if col not in other_answers.varnames
]
other_answers._assert_allclose(instance.trace.drop(*strip_out,
inplace=False),
rtol=RTOL,
atol=ATOL)
def setUp(self):
data = south()
data['n_samples'] = 0
with open(FULL_PATH + '/data/psrf_noburn.json', 'r') as noburn:
self.noburn = json.load(noburn)
with open(FULL_PATH + '/data/psrf_brooks.json', 'r') as brooks:
self.known_brooks = json.load(brooks)
with open(FULL_PATH + '/data/psrf_gr.json', 'r') as gr:
self.known_gr = json.load(gr)
np.random.seed(TEST_SEED)
self.trace = Trace.from_csv(FULL_PATH + '/data/south_mvcm_5000',
multi=True)
self.mockmodel = Hashmap(trace=self.trace)
def setUp(self):
self.a = {chr(i + 97): list(range(10)) for i in range(5)}
self.t = Trace(**self.a)
self.mt = Trace(self.a, self.a, self.a)
self.real_mt = Trace.from_csv(FULL_PATH + r'/data/south_mvcm_5000',
multi=True)
self.real_singles = [
Trace.from_csv(FULL_PATH +
r'/data/south_mvcm_5000_{}.csv'.format(i))
for i in range(4)
]
def test_from_df(self):
df = self.t.to_df()
new_trace = Trace.from_df(df)
assert new_trace == self.t
new_mt = Trace.from_df((df, df, df))
assert new_mt == self.t
def setUp(self):
super(Test_SMASE, self).build_self()
self.cls = M.SMASE
self.inputs['n_samples'] = 0
self.instance = self.cls(**self.inputs)
self.answer_trace = Trace.from_csv(FULL_PATH + '/data/smase.csv')
def setUp(self):
data = south()
data['n_samples'] = 0
np.random.seed(TEST_SEED)
self.trace = Trace.from_csv(FULL_PATH + '/data/south_mvcm_5000', multi=True)
def test_from_csv(self):
self.t.to_csv('./test_from_csv.csv')
new_t = Trace.from_csv('./test_from_csv.csv')
assert self.t == new_t
os.remove('./test_from_csv.csv')
def test_validate_names(self):
b = self.a.copy()
try:
bad_names = Trace(self.a, b, self.a, self.a)
except KeyError:
pass
def test_multi_roundtrips(self):
dfs = self.real_mt.to_df()
new = Trace.from_df(dfs)
new._assert_allclose(self.real_mt)
def test_ordering(self):
for ch, alone in zip(self.real_mt.chains, self.real_singles):
Trace(ch)._assert_allclose(alone)
def test_from_df(self):
df = self.t.to_df()
new_trace = Trace.from_df(df)
assert new_trace == self.t
new_mt = Trace.from_df((df, df, df))
assert new_mt == self.t
def setUp(self):
data = south()
data['n_samples'] = 0
np.random.seed(TEST_SEED)
self.trace = Trace.from_csv(FULL_PATH + '/data/south_mvcm_5000',
multi=True)
def setUp(self):
super(Test_Generic, self).build_self()
self.cls = M.Generic
self.inputs['n_samples'] = 0
instance = self.cls(**self.inputs)
self.answer_trace = Trace.from_csv(FULL_PATH + '/data/generic.csv')
def test_single_roundtrips(self):
source_from_file = self.real_singles[0]
from_df = Trace.from_df(source_from_file.to_df())
source_from_file._assert_allclose(from_df)
def test_multi_roundtrips(self):
dfs = self.real_mt.to_df()
new = Trace.from_df(dfs)
new._assert_allclose(self.real_mt)
def setUp(self):
super(Test_Generic, self).build_self()
self.cls = M.Generic
self.inputs['n_samples'] = 0
instance = self.cls(**self.inputs)
self.answer_trace = Trace.from_csv(FULL_PATH + '/data/generic.csv')
def test_from_csv(self):
self.t.to_csv('./test_from_csv.csv')
new_t = Trace.from_csv('./test_from_csv.csv')
assert self.t == new_t
os.remove('./test_from_csv.csv')
def test_single_roundtrips(self):
source_from_file = self.real_singles[0]
from_df = Trace.from_df(source_from_file.to_df())
source_from_file._assert_allclose(from_df)
def __init__(
self,
#data parameters
Y,
X,
coordinates,
n_samples=1000,
n_jobs=1,
priors=None,
configs=None,
starting_values=None,
extra_traced_params=None,
dmetric='euclidean',
correlation_function=nexp,
verbose=False,
center=True,
rescale_dists=True):
if center:
X = verify_center(X)
X = verify_covariates(X)
N, p = X.shape
Xs = X
X = explode(X)
self.state = Hashmap(X=X, Y=Y, coordinates=coordinates)
self.traced_params = ['Betas', 'Mus', 'T', 'Phi', 'Tau2']
if extra_traced_params is not None:
self.traced_params.extend(extra_traced_params)
self.trace = Trace(**{param: [] for param in self.traced_params})
st = self.state
self.state.correlation_function = correlation_function
self.verbose = verbose
st.Y = Y
st.X = X
st.Xs = Xs
st.N = N
st.p = p
st._dmetric = dmetric
if isinstance(st._dmetric, str):
st.pwds = d.squareform(d.pdist(st.coordinates, metric=st._dmetric))
elif callable(st._dmetric):
st.pwds = st._dmetric(st.coordinates)
st.max_dist = st.pwds.max()
if rescale_dists:
st.pwds = st.pwds / st.max_dist
st._old_max = st.max_dist
st.max_dist = 1.
if configs is None:
configs = dict()
if priors is None:
priors = dict()
if starting_values is None:
starting_values = dict()
self._setup_priors(**priors)
self._setup_starting_values(**starting_values)
self._setup_configs(**configs)
self._verbose = verbose
self.cycles = 0
if n_samples > 0:
try:
self.sample(n_samples, n_jobs=n_jobs)
except (np.linalg.LinAlgError, ValueError) as e:
Warn('Encountered the following LinAlgError. '
'Model will return for debugging. \n {}'.format(e))
class Test_Trace(ut.TestCase):
def setUp(self):
self.a = {chr(i+97):list(range(10)) for i in range(5)}
self.t = Trace(**self.a)
self.mt = Trace(self.a,self.a,self.a)
self.real_mt = Trace.from_csv(FULL_PATH + r'/data/south_mvcm_5000', multi=True)
self.real_singles = [Trace.from_csv(FULL_PATH + r'/data/south_mvcm_5000_{}.csv'
.format(i)) for i in range(4)]
def test_validate_names(self):
b = self.a.copy()
try:
bad_names = Trace(self.a,b,self.a,self.a)
except KeyError:
pass
# tests
def test_slicing(self):
t = self.t
mt = self.mt
#1index
assert t[1] == {'a':1, 'b':1, 'c':1, 'd':1, 'e':1}
assert mt[6] == [{k:6 for k in ['a','b', 'c', 'd', 'e']}]*3
assert t[-1] =={k:9 for k in ['a','b', 'c', 'd', 'e']}
assert mt[-1] == [{k:9 for k in ['a','b', 'c', 'd', 'e']}]*3
assert t[2:5] == {k:list(range(2,5)) for k in ['a','b', 'c', 'd', 'e']}
assert mt[8:] == [ {k:list(range(8,10)) for k in ['a','b', 'c', 'd', 'e'] }] * 3
assert t[-4::2] == {k:[6,8] for k in ['a','b', 'c', 'd', 'e']}
assert (t['a'] == list(range(10))).all()
assert (mt['a'] == [list(range(10))]*3).all()
assert t[['a','b']] == {'a':list(range(10)), 'b':list(range(10))}
assert mt[['a','b']] == [{'a':list(range(10)), 'b':list(range(10))}]*3
#2index
assert t['a', 1] == 1
assert t[['a', 'b'], 1] == {'a':1, 'b':1}
assert (mt['e', 5] == [5]*3).all()
assert mt[['d', 'e'], 8:] == [{'d':[8,9], 'e':[8,9]}]*3
assert (t[0, 'a'] == list(range(10))).all()
assert t[0, ['a', 'b']] == {'a':list(range(10)), 'b':list(range(10))}
try:
t[1, ['a','c']]
raise Exception('This did not raise an exception within the slicer!')
except IndexError:
pass
assert mt[1:, ['a','c']] == [{'a':list(range(10)), 'c':list(range(10))}] * 2
assert (mt[2, 'a'] == list(range(10))).all()
assert t[0,-1] == {k:9 for k in ['a', 'b', 'c', 'd', 'e']}
assert t[0,:] == {k:list(range(10)) for k in ['a', 'b', 'c', 'd', 'e']}
assert mt[:, -1:-4:-1] == [{k:[9,8,7] for k in ['a', 'b', 'c', 'd', 'e']}]*3
#3index
assert t[0, 'a', -1] == 9
assert t[0, ['a','b'],-3::2] == {'a':[7,9], 'b':[7,9]}
assert t[0, : ,-1] == {k:9 for k in ['a','b','c','d','e']}
try:
t[1, 'a', -1]
raise Exception('this did not raise an exception when it should have')
except IndexError:
pass
assert (mt[1:, 'a', -1] == [9]*2).all()
assert mt[1:, ['a','b'], -2:] == [{'a':[8,9], 'b':[8,9]}]*2
assert (mt[2, 'a', 5::2] == [5,7,9]).all()
assert (mt[1:, 'a', -5::2] == [[5,7,9]]*2).all()
assert (mt[:, 'a', -5::2] == [[5,7,9]]*3).all()
assert mt[2, :, :] == {k:list(range(10)) for k in ['a','b','c','d','e']}
assert mt[:,:,:] == mt.chains
assert mt[:,:,:] is not mt.chains
def test_to_df(self):
df = self.t.to_df()
df2 = pd.DataFrame.from_dict(self.t.chains[0])
np.testing.assert_array_equal(df.values, df2.values)
mtdf = self.mt.to_df()
mtdf2 = [pd.DataFrame.from_dict(chain) for chain in self.mt.chains]
for i in range(len(mtdf2)):
np.testing.assert_array_equal(mtdf[i].values, mtdf2[i].values)
def test_from_df(self):
df = self.t.to_df()
new_trace = Trace.from_df(df)
assert new_trace == self.t
new_mt = Trace.from_df((df, df, df))
assert new_mt == self.t
def test_to_csv(self):
df = self.t.to_df()
self.t.to_csv('./test_to_csv.csv')
new_df = pd.read_csv('./test_to_csv.csv')
np.testing.assert_allclose(df.values, new_df.values,
rtol=RTOL, atol=ATOL)
os.remove('./test_to_csv.csv')
def test_from_csv(self):
self.t.to_csv('./test_from_csv.csv')
new_t = Trace.from_csv('./test_from_csv.csv')
assert self.t == new_t
os.remove('./test_from_csv.csv')
def test_single_roundtrips(self):
source_from_file = self.real_singles[0]
from_df = Trace.from_df(source_from_file.to_df())
source_from_file._assert_allclose(from_df)
def test_ordering(self):
for ch, alone in zip(self.real_mt.chains, self.real_singles):
Trace(ch)._assert_allclose(alone)
def test_multi_roundtrips(self):
dfs = self.real_mt.to_df()
new = Trace.from_df(dfs)
new._assert_allclose(self.real_mt)
@ut.skip
def test_from_pymc3(self):
raise NotImplementedError
@ut.skip
def test_plot(self):
try:
import matplotlib as mpl
mpl.use('Agg')
self.t.plot()
except:
raise Exception('Single trace plotting failed!')
try:
import matplotlib as mpl
mpl.use('Agg')
self.mt.plot()
except:
raise Exception('Multi-chain trace plotting failed!')
def setUp(self):
super(Test_SESE, self).build_self()
self.cls = M.SESE
self.inputs['n_samples'] = 0
self.instance = self.cls(**self.inputs)
self.answer_trace = Trace.from_csv(FULL_PATH + '/data/sese.csv')
class Test_Trace(ut.TestCase):
def setUp(self):
self.a = {chr(i + 97): list(range(10)) for i in range(5)}
self.t = Trace(**self.a)
self.mt = Trace(self.a, self.a, self.a)
self.real_mt = Trace.from_csv(FULL_PATH + r'/data/south_mvcm_5000',
multi=True)
self.real_singles = [
Trace.from_csv(FULL_PATH +
r'/data/south_mvcm_5000_{}.csv'.format(i))
for i in range(4)
]
def test_validate_names(self):
b = self.a.copy()
try:
bad_names = Trace(self.a, b, self.a, self.a)
except KeyError:
pass
# tests
def test_slicing(self):
t = self.t
mt = self.mt
#1index
assert t[1] == {'a': 1, 'b': 1, 'c': 1, 'd': 1, 'e': 1}
assert mt[6] == [{k: 6 for k in ['a', 'b', 'c', 'd', 'e']}] * 3
assert t[-1] == {k: 9 for k in ['a', 'b', 'c', 'd', 'e']}
assert mt[-1] == [{k: 9 for k in ['a', 'b', 'c', 'd', 'e']}] * 3
assert t[2:5] == {
k: list(range(2, 5))
for k in ['a', 'b', 'c', 'd', 'e']
}
assert mt[8:] == [{
k: list(range(8, 10))
for k in ['a', 'b', 'c', 'd', 'e']
}] * 3
assert t[-4::2] == {k: [6, 8] for k in ['a', 'b', 'c', 'd', 'e']}
assert (t['a'] == list(range(10))).all()
assert (mt['a'] == [list(range(10))] * 3).all()
assert t[['a', 'b']] == {'a': list(range(10)), 'b': list(range(10))}
assert mt[['a', 'b']] == [{
'a': list(range(10)),
'b': list(range(10))
}] * 3
#2index
assert t['a', 1] == 1
assert t[['a', 'b'], 1] == {'a': 1, 'b': 1}
assert (mt['e', 5] == [5] * 3).all()
assert mt[['d', 'e'], 8:] == [{'d': [8, 9], 'e': [8, 9]}] * 3
assert (t[0, 'a'] == list(range(10))).all()
assert t[0, ['a', 'b']] == {'a': list(range(10)), 'b': list(range(10))}
try:
t[1, ['a', 'c']]
raise Exception(
'This did not raise an exception within the slicer!')
except IndexError:
pass
assert mt[1:, ['a', 'c']] == [{
'a': list(range(10)),
'c': list(range(10))
}] * 2
assert (mt[2, 'a'] == list(range(10))).all()
assert t[0, -1] == {k: 9 for k in ['a', 'b', 'c', 'd', 'e']}
assert t[0, :] == {
k: list(range(10))
for k in ['a', 'b', 'c', 'd', 'e']
}
assert mt[:, -1:-4:-1] == [{
k: [9, 8, 7]
for k in ['a', 'b', 'c', 'd', 'e']
}] * 3
#3index
assert t[0, 'a', -1] == 9
assert t[0, ['a', 'b'], -3::2] == {'a': [7, 9], 'b': [7, 9]}
assert t[0, :, -1] == {k: 9 for k in ['a', 'b', 'c', 'd', 'e']}
try:
t[1, 'a', -1]
raise Exception(
'this did not raise an exception when it should have')
except IndexError:
pass
assert (mt[1:, 'a', -1] == [9] * 2).all()
assert mt[1:, ['a', 'b'], -2:] == [{'a': [8, 9], 'b': [8, 9]}] * 2
assert (mt[2, 'a', 5::2] == [5, 7, 9]).all()
assert (mt[1:, 'a', -5::2] == [[5, 7, 9]] * 2).all()
assert (mt[:, 'a', -5::2] == [[5, 7, 9]] * 3).all()
assert mt[2, :, :] == {
k: list(range(10))
for k in ['a', 'b', 'c', 'd', 'e']
}
assert mt[:, :, :] == mt.chains
assert mt[:, :, :] is not mt.chains
def test_to_df(self):
df = self.t.to_df()
df2 = pd.DataFrame.from_dict(self.t.chains[0])
np.testing.assert_array_equal(df.values, df2.values)
mtdf = self.mt.to_df()
mtdf2 = [pd.DataFrame.from_dict(chain) for chain in self.mt.chains]
for i in range(len(mtdf2)):
np.testing.assert_array_equal(mtdf[i].values, mtdf2[i].values)
def test_from_df(self):
df = self.t.to_df()
new_trace = Trace.from_df(df)
assert new_trace == self.t
new_mt = Trace.from_df((df, df, df))
assert new_mt == self.t
def test_to_csv(self):
df = self.t.to_df()
self.t.to_csv('./test_to_csv.csv')
new_df = pd.read_csv('./test_to_csv.csv')
np.testing.assert_allclose(df.values,
new_df.values,
rtol=RTOL,
atol=ATOL)
os.remove('./test_to_csv.csv')
def test_from_csv(self):
self.t.to_csv('./test_from_csv.csv')
new_t = Trace.from_csv('./test_from_csv.csv')
assert self.t == new_t
os.remove('./test_from_csv.csv')
def test_single_roundtrips(self):
source_from_file = self.real_singles[0]
from_df = Trace.from_df(source_from_file.to_df())
source_from_file._assert_allclose(from_df)
def test_ordering(self):
for ch, alone in zip(self.real_mt.chains, self.real_singles):
Trace(ch)._assert_allclose(alone)
def test_multi_roundtrips(self):
dfs = self.real_mt.to_df()
new = Trace.from_df(dfs)
new._assert_allclose(self.real_mt)
@ut.skip
def test_from_pymc3(self):
raise NotImplementedError
@ut.skip
def test_plot(self):
try:
import matplotlib as mpl
mpl.use('Agg')
self.t.plot()
except:
raise Exception('Single trace plotting failed!')
try:
import matplotlib as mpl
mpl.use('Agg')
self.mt.plot()
except:
raise Exception('Multi-chain trace plotting failed!')