def set_sample(self):
sample_cols = tuple(
[self.__dict__[x] for x in self.sample_cols_labels])
self.sample = flag_sample(self.df, *sample_cols)
if self.nosingles and self.fe_name:
self.sample &= flag_nonsingletons(self.df, self.fe_name,
self.sample)
sample_vars = set_sample(self.df, self.sample, sample_cols)
self.__dict__.update(dict(zip(self.sample_store_labels, sample_vars)))
self.x = force_df(self.x)
self.y = self.y.squeeze()
# Force regression variables to float64
for var in self.vars_in_reg:
self.__dict__[var] = self.__dict__[var].astype(np.float64)
# Demean or add constant
if self.fe_name is not None:
self._demean_sample()
elif self.addcons:
_cons = np.ones(self.y.shape[0])
x = self.__dict__[self.add_constant_to]
if x.empty:
x = pd.DataFrame(_cons, columns=['_cons'], index=self.y.index)
else:
x['_cons'] = _cons
self.__dict__[self.add_constant_to] = x
# Re-weight sample
if self.AWT is not None:
self._weight_sample()
def set_sample(self):
sample_cols = tuple(
[self.__dict__[x] for x in self.sample_cols_labels])
self.sample = flag_sample(self.df, *sample_cols)
if self.nosingles and self.fe_name:
self.sample &= flag_nonsingletons(self.df, self.fe_name,
self.sample)
sample_vars = set_sample(self.df, self.sample, sample_cols)
self.__dict__.update(dict(zip(self.sample_store_labels, sample_vars)))
self.x = force_df(self.x)
self.y = self.y.squeeze()
# Force regression variables to float64
for var in self.vars_in_reg:
self.__dict__[var] = self.__dict__[var].astype(np.float64)
# Demean or add constant
if self.fe_name is not None:
self._demean_sample()
elif self.addcons:
_cons = np.ones(self.y.shape[0])
x = self.__dict__[self.add_constant_to]
if x.empty:
x = pd.DataFrame(_cons, columns=['_cons'], index=self.y.index)
else:
x['_cons'] = _cons
self.__dict__[self.add_constant_to] = x
# Re-weight sample
if self.AWT is not None:
self._weight_sample()
def test_multiidx(self):
idx = pd.MultiIndex.from_tuples([('a', 1), ('b', 7), ('w', 99)],
names=['dingle', 'dangle'])
expected = self.sdf
expected.index = idx
s = self.s
s.index = idx
result = force_df(s)
assert_frame_equal(expected, result)
def test_multiidx(self):
idx = pd.MultiIndex.from_tuples([('a', 1), ('b', 7), ('w', 99)],
names=['dingle', 'dangle'])
expected = self.sdf
expected.index = idx
s = self.s
s.index = idx
result = force_df(s)
assert_frame_equal(expected, result)
def _demean_guts(A, args):
for df in args:
# Ignore empty `df` (e.g. empty list of exogenous included regressors)
if df is None or df.empty:
yield df
else:
group_name = A.name
mean = df.groupby(A).mean()
large_mean = force_df(A).join(mean, on=group_name).drop(group_name,
axis=1)
if df.ndim == 1:
large_mean = large_mean.squeeze()
demeaned = df - large_mean
yield demeaned
def _demean_guts(A, args): #noqa
for df in args:
# Ignore empty `df` (e.g. empty list of exogenous included regressors)
if df is None or df.empty:
yield df
else:
group_name = A.name
mean = df.groupby(A).mean()
large_mean = force_df(A).join(mean, on=group_name).drop(group_name,
axis=1)
if df.ndim == 1:
large_mean = large_mean.squeeze()
demeaned = df - large_mean
yield demeaned
def _demean(A, df):
""" Demean a matrix/DataFrame within group `A` """
# Ignore empty `df` (e.g. empty list of exogenous included regressors)
if df is None or df.empty:
return df
else:
group_name = A.name
mean = df.groupby(A).mean()
large_mean = force_df(A).join(mean, on=group_name).drop(group_name,
axis=1)
if df.ndim == 1:
large_mean = large_mean.squeeze()
demeaned = df - large_mean
return demeaned
def _demean(A, df):
""" Demean a matrix/DataFrame within group `A` """
# Ignore empty `df` (e.g. empty list of exogenous included regressors)
if df is None or df.empty:
return df
else:
group_name = A.name
mean = df.groupby(A).mean()
large_mean = force_df(A).join(mean, on=group_name).drop(group_name,
axis=1)
if df.ndim == 1:
large_mean = large_mean.squeeze()
demeaned = df - large_mean
return demeaned
def test_list(self):
with pytest.raises(ValueError):
force_df([1, 2, 3])
def test_array(self):
force_df(np.arange(3))
def test_list(self):
force_df([1, 2, 3])
def test_simple(self):
expected = pd.DataFrame(self.s)
result = force_df(self.s)
assert_frame_equal(expected, result)
def test_passthrough(self):
df = self.df
assert_frame_equal(df, force_df(df))
def test_simple(self):
expected = pd.DataFrame(self.s)
result = force_df(self.s)
assert_frame_equal(expected, result)
def test_passthrough(self):
df = self.df
assert_frame_equal(df, force_df(df))
def test_array(self):
with pytest.raises(ValueError):
force_df(np.arange(3))
def test_array(self):
with pytest.raises(ValueError):
force_df(np.arange(3))
def test_list(self):
with pytest.raises(ValueError):
force_df([1, 2, 3])
