def test_concat_sort(data):
a = concat([data.df1, data.df2], 1)
b = concat([data.df1, data.df2, data.s], 1)
c = concat([data.df1, data.df2, data.s], 1, sort=True)
d = concat([data.df2, data.df1, data.s], 1, sort=False)
assert list(a.columns) == ['A', 'B', 'C']
assert list(b.columns) == ['A', 'B', 'C', 'D']
assert list(c.columns) == ['A', 'B', 'C', 'D']
assert list(d.columns) == ['B', 'C', 'A', 'D']
def test_concat_sort(data):
a = concat([data.df1, data.df2], 1)
b = concat([data.df1, data.df2, data.s], 1)
c = concat([data.df1, data.df2, data.s], 1, sort=True)
d = concat([data.df2, data.df1, data.s], 1, sort=False)
assert list(a.columns) == ["A", "B", "C"]
assert list(b.columns) == ["A", "B", "C", "D"]
assert list(c.columns) == ["A", "B", "C", "D"]
assert list(d.columns) == ["B", "C", "A", "D"]
def fit(request):
method = request.param
data = generate_simultaneous_data()
if "ols" in method or "sur" in method:
mod = SUR
for key in data:
temp = data[key]
temp["exog"] = concat([temp["exog"], temp["endog"]], 1)
del temp["endog"]
del temp["instruments"]
else:
mod = IV3SLS
if "ols" in method or "2sls" in method:
fit_method = "ols"
else:
fit_method = "gls"
mod = mod(data)
iterate = "ireg3" in method
stata = results[method]
debiased = method in ("ols", "2sls")
kwargs = {}
decimal = 3 if "ireg3" in method else 5
rtol = 10**-decimal
res = mod.fit(
cov_type="unadjusted",
method=fit_method,
debiased=debiased,
iterate=iterate,
**kwargs,
)
return stata, res, rtol
def _get_series_property(self, name: str) -> DataFrame:
out: List[Tuple[str, Series]] = [(k, getattr(v, name))
for k, v in self._results.items()]
cols = [v[0] for v in out]
values = concat([v[1] for v in out], axis=1)
values.columns = cols
return values
def fit(request):
method = request.param
data = generate_simultaneous_data()
if 'ols' in method or 'sur' in method:
mod = SUR
for key in data:
temp = data[key]
temp['exog'] = concat([temp['exog'], temp['endog']], 1)
del temp['endog']
del temp['instruments']
else:
mod = IV3SLS
if 'ols' in method or '2sls' in method:
fit_method = 'ols'
else:
fit_method = 'gls'
mod = mod(data)
iterate = 'ireg3' in method
stata = results[method]
debiased = method in ('ols', '2sls')
kwargs = {}
decimal = 2 if 'ireg3' in method else 5
rtol = 10**-decimal
res = mod.fit(cov_type='unadjusted',
method=fit_method,
debiased=debiased,
iterate=iterate,
**kwargs)
return stata, res, rtol
def data():
premia = np.array([.1, .1, .1])
out = generate_data(nportfolio=10,
output='pandas',
alpha=True,
premia=premia)
out['joined'] = concat([out.factors, out.portfolios], 1)
return out
def test_formula_equivalence_weights(data):
weights = AttrDict()
eqn_copy = AttrDict()
for key in data.eqns:
eqn = {k: v for k, v in data.eqns[key].items()}
nobs = eqn["dependent"].shape[0]
w = np.random.chisquare(2, (nobs, 1)) / 2
weights[key] = w
eqn["weights"] = w
eqn_copy[key] = eqn
mod = IVSystemGMM(eqn_copy, weight_type="unadjusted")
df = []
formulas = {}
for i, key in enumerate(data.eqns):
eqn = data.eqns[key]
dep = eqn.dependent
ex = eqn.exog
en = eqn.endog
instr = eqn.instruments
dep = DataFrame(dep, columns=["dep_{0}".format(i)])
has_const = False
if np.any(np.all(ex == 1, 0)):
ex = ex[:, 1:]
has_const = True
ex = DataFrame(
ex,
columns=["ex_{0}_{1}".format(i, j) for j in range(ex.shape[1])])
en = DataFrame(
en,
columns=["en_{0}_{1}".format(i, j) for j in range(en.shape[1])])
instr = DataFrame(
instr,
columns=["instr_{0}_{1}".format(i, j) for j in range(ex.shape[1])])
fmla = "".join(dep.columns) + " ~ "
if has_const:
fmla += " 1 + "
fmla += " + ".join(ex.columns) + " + ["
fmla += " + ".join(en.columns) + " ~ "
fmla += " + ".join(instr.columns) + " ] "
formulas[key] = fmla
df.extend([dep, ex, en, instr])
df = concat(df, 1)
formula_mod = IVSystemGMM.from_formula(formulas,
df,
weights=weights,
weight_type="unadjusted")
res = mod.fit(cov_type="unadjusted")
formula_res = formula_mod.fit(cov_type="unadjusted")
assert_allclose(res.params, formula_res.params)
def test_formula_equivalence(data):
mod = IVSystemGMM(data.eqns, weight_type="unadjusted")
formula = []
df = []
for i, key in enumerate(data.eqns):
eqn = data.eqns[key]
dep = eqn.dependent
ex = eqn.exog
en = eqn.endog
instr = eqn.instruments
dep = DataFrame(dep, columns=["dep_{0}".format(i)])
has_const = False
if np.any(np.all(ex == 1, 0)):
ex = ex[:, 1:]
has_const = True
ex = DataFrame(
ex,
columns=["ex_{0}_{1}".format(i, j) for j in range(ex.shape[1])])
en = DataFrame(
en,
columns=["en_{0}_{1}".format(i, j) for j in range(en.shape[1])])
instr = DataFrame(
instr,
columns=["instr_{0}_{1}".format(i, j) for j in range(ex.shape[1])])
fmla = "".join(dep.columns) + " ~ "
if has_const:
fmla += " 1 + "
fmla += " + ".join(ex.columns) + " + ["
fmla += " + ".join(en.columns) + " ~ "
fmla += " + ".join(instr.columns) + " ] "
formula.append(fmla)
df.extend([dep, ex, en, instr])
formulas = {}
for i, f in enumerate(formula):
formulas["eq{0}".format(i)] = f
df = concat(df, 1)
formula_mod = IVSystemGMM.from_formula(formulas,
df,
weight_type="unadjusted")
res = mod.fit(cov_type="unadjusted")
formula_res = formula_mod.fit(cov_type="unadjusted")
assert_allclose(res.params, formula_res.params)
def test_predict_formula_function(data, model_and_func):
model, func = model_and_func
fmla = 'y ~ 1 + sigmoid(x3) + x4 + [x1 + x2 ~ z1 + z2 + z3] + np.exp(x5)'
mod = model.from_formula(fmla, data)
res = mod.fit()
exog = [data[['Intercept']], sigmoid(data[['x3']]), data[['x4']],
np.exp(data[['x5']])]
exog = concat(exog, 1)
endog = data[['x1', 'x2']]
pred = res.predict(exog, endog)
pred2 = res.predict(data=data)
assert_frame_equal(pred, pred2)
assert_allclose(res.fitted_values, pred)
res2 = func(fmla, data).fit()
pred3 = res2.predict(exog, endog)
pred4 = res2.predict(data=data)
assert_frame_equal(pred, pred3)
assert_frame_equal(pred, pred4)
def test_formula_function(data, model_and_func):
model, func = model_and_func
fmla = 'y ~ 1 + sigmoid(x3) + x4 + [x1 + x2 ~ z1 + z2 + z3] + np.exp(x5)'
mod = model.from_formula(fmla, data)
res = mod.fit()
dep = data.y
exog = [data[['Intercept']], sigmoid(data[['x3']]), data[['x4']],
np.exp(data[['x5']])]
exog = concat(exog, 1)
endog = data[['x1', 'x2']]
instr = data[['z1', 'z2', 'z3']]
mod = model(dep, exog, endog, instr)
res2 = mod.fit()
assert_equal(res.params.values, res2.params.values)
res3 = func(fmla, data).fit()
assert_equal(res.params.values, res3.params.values)
with pytest.raises(ValueError):
res2.predict(data=data)
def test_formula_function(data, model_and_func):
model, func = model_and_func
fmla = "y ~ 1 + sigmoid(x3) + x4 + [x1 + x2 ~ z1 + z2 + z3] + np.exp(x5)"
mod = model.from_formula(fmla, data)
res = mod.fit()
dep = data.y
exog = [
data[["Intercept"]],
sigmoid(data[["x3"]]),
data[["x4"]],
np.exp(data[["x5"]]),
]
exog = concat(exog, 1)
endog = data[["x1", "x2"]]
instr = data[["z1", "z2", "z3"]]
mod = model(dep, exog, endog, instr)
res2 = mod.fit()
assert_equal(res.params.values, res2.params.values)
res3 = func(fmla, data).fit()
assert_equal(res.params.values, res3.params.values)
with pytest.raises(ValueError):
res2.predict(data=data)
def test_formula_equivalence(data):
mod = IVSystemGMM(data.eqns, weight_type='unadjusted')
formula = []
df = []
for i, key in enumerate(data.eqns):
eqn = data.eqns[key]
dep = eqn.dependent
ex = eqn.exog
en = eqn.endog
instr = eqn.instruments
dep = DataFrame(dep, columns=['dep_{0}'.format(i)])
has_const = False
if np.any(np.all(ex == 1, 0)):
ex = ex[:, 1:]
has_const = True
ex = DataFrame(ex, columns=['ex_{0}_{1}'.format(i, j) for j in range(ex.shape[1])])
en = DataFrame(en, columns=['en_{0}_{1}'.format(i, j) for j in range(en.shape[1])])
instr = DataFrame(instr, columns=['instr_{0}_{1}'.format(i, j)
for j in range(ex.shape[1])])
fmla = ''.join(dep.columns) + ' ~ '
if has_const:
fmla += ' 1 + '
fmla += ' + '.join(ex.columns) + ' + ['
fmla += ' + '.join(en.columns) + ' ~ '
fmla += ' + '.join(instr.columns) + ' ] '
formula.append(fmla)
df.extend([dep, ex, en, instr])
from collections import OrderedDict
formulas = OrderedDict()
for i, f in enumerate(formula):
formulas['eq{0}'.format(i)] = f
df = concat(df, 1)
formula_mod = IVSystemGMM.from_formula(formulas, df, weight_type='unadjusted')
res = mod.fit(cov_type='unadjusted')
formula_res = formula_mod.fit(cov_type='unadjusted')
assert_allclose(res.params, formula_res.params)
w = w / w.mean()
items = ['x' + str(i) for i in range(1, k + 1)]
items = ['intercept'] + items
major = pd.date_range('12-31-1999', periods=t, freq='A-DEC')
minor = ['firm.' + str(i) for i in range(1, n + 1)]
x = panel_to_frame(x, items, major, minor, swap=True)
y = panel_to_frame(y[None, :], ['y'], major, minor, swap=True)
w = panel_to_frame(w[None, :], ['w'], major, minor, swap=True)
x = PanelData(x)
y = PanelData(y)
w = PanelData(w)
z = concat([x.dataframe, y.dataframe, w.dataframe], 1)
final_index = pd.MultiIndex.from_product([minor, major])
final_index.levels[0].name = 'firm'
z = z.reindex(final_index)
z.index.levels[0].name = 'firm'
z.index.levels[1].name = 'time'
z = z.reset_index()
z['firm_id'] = z.firm.astype('category')
z['firm_id'] = z.firm_id.cat.codes
vars = ['y', 'x1', 'x2', 'x3', 'x4', 'x5']
missing = 0.05
for v in vars:
locs = np.random.choice(n * t, int(n * t * missing))
temp = z[v].copy()
def _get_series_property(self, name):
out = ([(k, getattr(v, name)) for k, v in self._results.items()])
cols = [v[0] for v in out]
values = concat([v[1] for v in out], 1)
values.columns = cols
return values
if np.any(locs):
dep.flat[locs] = np.nan
exog = missing_data[key]['exog']
locs = np.where(np.random.random_sample(np.prod(exog.shape)) < 0.02)[0]
if np.any(locs):
exog.flat[locs] = np.nan
out = []
for i, dataset in enumerate((basic_data, common_data, missing_data)):
base = 'mod_{0}'.format(i)
for j, key in enumerate(dataset):
dep = dataset[key]['dependent']
dep = pd.DataFrame(dep, columns=[base + '_y_{0}'.format(j)])
dataset[key]['dependent'] = dep
exog = dataset[key]['exog'][:, 1:]
exog_cols = [
base + '_x_{0}{1}'.format(j, k) for k in range(exog.shape[1])
]
exog = pd.DataFrame(exog, columns=exog_cols)
exog = exog.copy()
exog['cons'] = 1.0
dataset[key]['exog'] = exog
if i != 1 or j == 0:
out.extend([dep, exog])
else:
out.extend([dep])
if __name__ == '__main__':
df = concat(out, 1)
df.to_stata('simulated-sur.dta')
def generate_panel_data(
nentity: int = 971,
ntime: int = 7,
nexog: int = 5,
const: bool = False,
missing: float = 0,
other_effects: int = 2,
ncats: Union[int, List[int]] = 4,
rng: Optional[np.random.RandomState] = None,
) -> PanelModelData:
"""
Parameters
----------
nentity : int, default 971
The number of entities in the panel.
ntime : int, default 7
The number of time periods in the panel.
nexog : int, default 5
The number of explanatory variables in the dataset.
const : bool, default False
Flag indicating that the model should include a constant.
missing : float, default 0
The percentage of values that are missing. Should be between 0 and 100.
other_effects : int, default 2
The number of other effects generated.
ncats : Union[int, Sequence[int]], default 4
The number of categories to use in other_effects and variance
clusters. If list-like, then it must have as many elements
as other_effects.
rng : RandomState, default None
A NumPy RandomState instance. If not provided, one is initialized
using a fixed seed.
Returns
-------
PanelModelData
A namedtuple derived class containing 4 DataFrames:
* `data` - A simulated data with variables y and x# for # in 0,...,4.
If const is True, then also contains a column named const.
* `weights` - Simulated non-negative weights.
* `other_effects` - Simulated effects.
* `clusters` - Simulated data to use in clustered covariance estimation.
"""
if rng is None:
rng = np.random.RandomState(
[
0xA14E2429,
0x448D2E51,
0x91B558E7,
0x6A3F5CD2,
0x22B43ABB,
0xE746C92D,
0xCE691A7D,
0x66746EE7,
]
)
n, t, k = nentity, ntime, nexog
k += int(const)
x = rng.standard_normal((k, t, n))
beta = np.arange(1, k + 1)[:, None, None] / k
y = (
(x * beta).sum(0)
+ rng.standard_normal((t, n))
+ 2 * rng.standard_normal((1, n))
)
w = rng.chisquare(5, (t, n)) / 5
c = None
cats = [f"cat.{i}" for i in range(other_effects)]
if other_effects:
if not isinstance(ncats, list):
ncats = [ncats] * other_effects
c = []
for i in range(other_effects):
nc = ncats[i]
c.append(rng.randint(0, nc, (1, t, n)))
c = np.concatenate(c, 0)
vcats = [f"varcat.{i}" for i in range(2)]
vc2 = np.ones((2, t, 1)) @ rng.randint(0, n // 2, (2, 1, n))
vc1 = vc2[[0]]
if const:
x[0] = 1.0
if missing > 0:
locs = rng.choice(n * t, int(n * t * missing))
y.flat[locs] = np.nan
locs = rng.choice(n * t * k, int(n * t * k * missing))
x.flat[locs] = np.nan
entities = [f"firm{i}" for i in range(n)]
time = date_range("1-1-1900", periods=t, freq="A-DEC")
var_names = [f"x{i}" for i in range(k)]
if const:
var_names[1:] = var_names[:-1]
var_names[0] = "const"
# y = DataFrame(y, index=time, columns=entities)
y_df = panel_to_frame(
y[None], items=["y"], major_axis=time, minor_axis=entities, swap=True
)
index = y_df.index
w_df = panel_to_frame(
w[None], items=["w"], major_axis=time, minor_axis=entities, swap=True
)
w_df = w_df.reindex(index)
x_df = panel_to_frame(
x, items=var_names, major_axis=time, minor_axis=entities, swap=True
)
x_df = x_df.reindex(index)
c_df = panel_to_frame(
c, items=cats, major_axis=time, minor_axis=entities, swap=True
)
other_eff = c_df.reindex(index)
vc1_df = panel_to_frame(
vc1, items=vcats[:1], major_axis=time, minor_axis=entities, swap=True
)
vc1_df = vc1_df.reindex(index)
vc2_df = panel_to_frame(
vc2, items=vcats, major_axis=time, minor_axis=entities, swap=True
)
vc2_df = vc2_df.reindex(index)
clusters = concat([vc1_df, vc2_df])
data = concat([y_df, x_df], axis=1)
return PanelModelData(data, w_df, other_eff, clusters)
def expand_categoricals(x, drop_first):
if x.shape[1] == 0:
return x
return concat([convert_columns(x[c], drop_first) for c in x.columns],
axis=1)
from linearmodels.tests.system._utility import generate_simultaneous_data
data = generate_simultaneous_data()
all_cols = []
out = []
for key in data:
eqn = data[key]
for key in ('exog', 'endog'):
vals = eqn[key]
for col in vals:
if col in all_cols:
continue
else:
out.append(vals[col])
all_cols.append(col)
out = concat(out, 1)
if 'const' in out:
out.pop('const')
out.to_stata('simulated-3sls.dta', write_index=False)
SEP = """
file open myfile using {outfile}, write append
file write myfile "!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! {method} !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!" _n
file close myfile
"""
# add , 2sls to get non GLS estimator
CMD = """
reg3 (dependent_0 dependent_1 dependent_2 exog_1 exog_2 exog_3 exog_4 exog_5) ///
(dependent_1 dependent_0 dependent_2 exog_1 exog_2 exog_3 exog_6 exog_7) ///
def data():
premia = np.array([0.1, 0.1, 0.1])
out = generate_data(nportfolio=10, output="pandas", alpha=True, premia=premia)
out["joined"] = concat([out.factors, out.portfolios], 1)
return out
def expand_categoricals(x: AnyPandas, drop_first: bool) -> AnyPandas:
if x.shape[1] == 0:
return x
return concat([convert_columns(x[c], drop_first) for c in x.columns],
axis=1)
def expand_categoricals(x: DataFrame, drop_first: bool) -> DataFrame:
return concat([convert_columns(x[c], drop_first) for c in x.columns],
axis=1)