def summary(self, title=None, alpha=.05):
"""
Summarize the #1lab_results of running MICE.
Parameters
-----------
title : string, optional
Title for the top table. If not None, then this replaces
the default title
alpha : float
Significance level for the confidence intervals
Returns
-------
smry : Summary instance
This holds the summary tables and text, which can be
printed or converted to various output formats.
"""
from statsmodels.iolib import summary2
from statsmodels.compat.collections import OrderedDict
smry = summary2.Summary()
float_format = "%8.3f"
info = OrderedDict()
info["Method:"] = "MICE"
info["Model:"] = self.model_class.__name__
info["Dependent variable:"] = self.endog_names
info["Sample size:"] = "%d" % self.model.data.data.shape[0]
info["Scale"] = "%.2f" % self.scale
info["Num. imputations"] = "%d" % len(self.model.results_list)
smry.add_dict(info, align='l', float_format=float_format)
param = summary2.summary_params(self, alpha=alpha)
param["FMI"] = self.frac_miss_info
smry.add_df(param, float_format=float_format)
smry.add_title(title=title, results=self)
return smry
def summary(self, title=None, alpha=.05):
"""
Summarize the results of running MICE.
Parameters
-----------
title : string, optional
Title for the top table. If not None, then this replaces
the default title
alpha : float
Significance level for the confidence intervals
Returns
-------
smry : Summary instance
This holds the summary tables and text, which can be
printed or converted to various output formats.
"""
from statsmodels.iolib import summary2
from statsmodels.compat.collections import OrderedDict
smry = summary2.Summary()
float_format = "%8.3f"
info = OrderedDict()
info["Method:"] = "MICE"
info["Model:"] = self.model_class.__name__
info["Dependent variable:"] = self.endog_names
info["Sample size:"] = "%d" % self.model.data.data.shape[0]
info["Scale"] = "%.2f" % self.scale
info["Num. imputations"] = "%d" % len(self.model.results_list)
smry.add_dict(info, align='l', float_format=float_format)
param = summary2.summary_params(self, alpha=alpha)
param["FMI"] = self.frac_miss_info
smry.add_df(param, float_format=float_format)
smry.add_title(title=title, results=self)
return smry
def summary(self, title=None, alpha=.05):
"""
Summarize the results of running multiple imputation.
Parameters
----------
title : str, optional
Title for the top table. If not None, then this replaces
the default title
alpha : float
Significance level for the confidence intervals
Returns
-------
smry : Summary instance
This holds the summary tables and text, which can be
printed or converted to various output formats.
"""
from statsmodels.iolib import summary2
smry = summary2.Summary()
float_format = "%8.3f"
info = {}
info["Method:"] = "MI"
info["Model:"] = self.mi.model.__name__
info["Dependent variable:"] = self._model.endog_names
info["Sample size:"] = "%d" % self.mi.imp.data.shape[0]
info["Num. imputations"] = "%d" % self.mi.nrep
smry.add_dict(info, align='l', float_format=float_format)
param = summary2.summary_params(self, alpha=alpha)
param["FMI"] = self.fmi
smry.add_df(param, float_format=float_format)
smry.add_title(title=title, results=self)
return smry
def col_params(result,
est_fmt='{:.3f}',
sig_fmt='({:.2f})',
est_scale=100.0,
sig_scale=1.0,
stars=True,
use_t=True,
xname=[],
**kwargs):
'''Stack coefficients and significance stats in single column
'''
# Extract parameters
res = summary_params(result, xname=xname)
# Format float
for col in res.columns[:2]:
res[col] = res[col].apply(lambda x: est_fmt.format(x * est_scale))
# tvalues in parentheses
if use_t:
res.iloc[:,
1] = [sig_fmt.format(x * sig_scale) for x in res.iloc[:, 2]]
else:
res.iloc[:,
1] = [sig_fmt.format(x * sig_scale) for x in res.iloc[:, 3]]
# Significance stars
if stars:
idx = res.iloc[:, 3] < .1
res.iloc[:, 0][idx] = res.iloc[:, 0][idx] + '*'
idx = res.iloc[:, 3] < .05
res.iloc[:, 0][idx] = res.iloc[:, 0][idx] + '*'
idx = res.iloc[:, 3] < .01
res.iloc[:, 0][idx] = res.iloc[:, 0][idx] + '*'
# Stack Coefs and Signif.
res = res.iloc[:, :2]
res = res.stack()
res = pd.DataFrame(res)
res.columns = [str(result.model.endog_names)]
return res
def _update_statsmodel_result_summary_after_cov_matrix_changed(result):
"""
Note: inplace
"""
# Create new param/stderr section of summary
new_param_stderr = summary_params(result)
new_table = SimpleTable(
new_param_stderr.values,
headers=list(new_param_stderr.columns),
stubs=list(new_param_stderr.index),
txt_fmt=fmt_params,
)
# Create summary object with param/stderr table replaced
summ = result.summary()
summ.tables[1] = new_table
# Assign summary method of result to return this summary object
result.summary = lambda: summ
# Repeat steps for summary2, which only requires df and not SimpleTable
summ2 = result.summary2()
summ2.tables[1] = new_param_stderr
result.summary2 = lambda: summ2
def produce_reg_df(model, model_name, panel, reg_type='ols'):
y, x = dmatrices(model, panel)
if reg_type == 'ols':
results = sm.OLS(y, x).fit()
estimates = summary_params(results)[['Coef.', 'Std.Err.', 'P>|t|']]
'''
White’s (1980) heteroskedasticity robust standard errors. Defined as
sqrt(diag(X.T X)^(-1)X.T diag(e_i^(2)) X(X.T X)^(-1) where e_i = resid[i]
HC0_se is a property. It is not evaluated until it is called. When it is
called the RegressionResults instance will then have another attribute
cov_HC0, which is the full heteroskedasticity consistent covariance matrix
and also het_scale, which is in this case just resid**2. HCCM matrices are
only appropriate for OLS.
Note: Delete the following two lines for 'regular' standard errors.
'''
estimates['Std.Err.'] = results.HC0_se
estimates['P>|t|'] = stats.t.sf(
np.abs(estimates['Coef.'] / estimates['Std.Err.']), results.nobs - 1) * 2
elif reg_type == 'probit':
model = sm.Probit(y, x)
results = model.fit()
margeffs = results.get_margeff()
estimates = pd.DataFrame(
[margeffs.margeff, margeffs.margeff_se, margeffs.pvalues],
index=['Coef.', 'Std.Err.', 'P>|t|'],
columns=model.exog_names[1:]).T
estimates = estimates.apply(
lambda x: ['{0:0.3f}'.format(i) for i in x])
estimates['Std.Err.'] = estimates['Std.Err.'].apply(
lambda x: '(' + str(x) + ')')
for i in range(len(estimates)):
estimates['Coef.'].iloc[i] = str(estimates['Coef.'].iloc[i]) + (
(float(estimates['P>|t|'].iloc[i]) <= 0.01) * '_3stars' +
(0.01 < float(estimates['P>|t|'].iloc[i]) <= 0.05) * '_2stars' +
(0.05 < float(estimates['P>|t|'].iloc[i]) <= 0.10) * '_1star' +
(0.1 < float(estimates['P>|t|'].iloc[i])) * ''
)
estimates['P>|t|'] = estimates['P>|t|'].apply(lambda x: '')
# Instead of inserting lines, just replace pvalues by linespace.
estimates = estimates.rename(columns={
'P>|t|': 'addlinespace'}
)
stacked_estimates = pd.DataFrame(
estimates.stack(), columns=[model_name])
if reg_type == 'ols':
stacked_model_stats = pd.DataFrame(
[results.nobs, results.rsquared_adj],
index=['Observations', 'R2'],
columns=[model_name])
elif reg_type == 'probit':
stacked_model_stats = pd.DataFrame(
[results.nobs, results.prsquared],
index=['Observations', 'R2'],
columns=[model_name])
stacked_model = stacked_estimates.append(stacked_model_stats)
return stacked_model
