def _add_point_estimate_ci(self, df: DataFrame):
df[CI_LOWER], df[CI_UPPER] = _zconfint_generic(
mean=df[POINT_ESTIMATE],
std_mean=np.sqrt(df[VARIANCE] / df[self._denominator]),
alpha=1 - self._interval_size,
alternative=TWO_SIDED)
return df
def conf_int(self, value=None, alpha=0.05, alternative="two-sided"):
"""
Confidence interval for probability that sample 1 has larger values
Confidence interval is for the shifted probability
P(x1 > x2) + 0.5 * P(x1 = x2) - value
Parameters
----------
value : float
Value, default 0, shifts the confidence interval,
e.g. ``value=0.5`` centers the confidence interval at zero.
alpha : float
Significance level for the confidence interval, coverage is
``1-alpha``
alternative : str
The alternative hypothesis, H1, has to be one of the following
* 'two-sided' : H1: ``prob - value`` not equal to 0.
* 'larger' : H1: ``prob - value > 0``
* 'smaller' : H1: ``prob - value < 0``
Returns
-------
lower : float or ndarray
Lower confidence limit. This is -inf for the one-sided alternative
"smaller".
upper : float or ndarray
Upper confidence limit. This is inf for the one-sided alternative
"larger".
"""
p0 = value
if p0 is None:
p0 = 0
diff = self.prob1 - p0
std_diff = np.sqrt(self.var / self.nobs)
if self.use_t is False:
return _zconfint_generic(diff, std_diff, alpha, alternative)
else:
return _tconfint_generic(diff, std_diff, self.df, alpha,
alternative)
def get_peruser_diff_zstat(df,
nobs_name,
mean_name,
std_name,
test_group_column='test_group',
test='TEST',
control='CONTROL',
alpha=0.05,
alternative='two-sided'):
'''confidence interval for the difference in means.
Similar to https://www.statsmodels.org/dev/generated/statsmodels.stats.weightstats.CompareMeans.zconfint_diff.html#statsmodels.stats.weightstats.CompareMeans.zconfint_diff
Example: pre = s1.build_and_run_sql(grouping=['test_group','user_id'], sql_template=TEMPLATE_USER_STATS)
get_user_agg_zstat(pre, 'units','buyers mean', 'buyers std')
Parameters
----------
alpha : float
significance level for the confidence interval, coverage is
``1-alpha``
alternative : string
This specifies the alternative hypothesis for the test that
corresponds to the confidence interval.
The alternative hypothesis, H1, has to be one of the following :
'two-sided': H1: difference in means not equal to value (default)
'larger' : H1: difference in means larger than value
'smaller' : H1: difference in means smaller than value
Returns
-------
diff, zstat, pvalue, confint : floats
'''
summary = df.set_index(test_group_column)#.T.loc[[nobs_name, mean_name, std_name],[test, control]].to_dict()
d1 = summary.loc[test, :]
d2 = summary.loc[control,:]
diff = d1[mean_name] - d2[mean_name]
std_diff = np.sqrt(d1[std_name]**2/(d1[nobs_name]-1) + d2[std_name]**2/(d2[nobs_name]-1)) #Assume unequal variance
confint = wstats._zconfint_generic(diff, std_diff, alpha=alpha, alternative=alternative)
zstat, pvalue = wstats._zstat_generic2(diff, std_diff, alternative)
return d1[mean_name], d2[mean_name], diff/d2[mean_name], confint/d2[mean_name], zstat, pvalue
def _ci(self, row, alpha_column: str) -> Tuple[float, float]:
return _zconfint_generic(mean=row[DIFFERENCE],
std_mean=row[STD_ERR],
alpha=row[alpha_column],
alternative=row[PREFERENCE])