Skip to content

Commit

Permalink
Modified Ci_corr to avoid if statements in test_corr
Browse files Browse the repository at this point in the history
  • Loading branch information
@j-grana6
j-grana6 committed Aug 9, 2012
1 parent fb112f9 commit 5d081c9
Showing 1 changed file with 51 additions and 57 deletions.
108 changes: 51 additions & 57 deletions statsmodels/emplike/descriptive.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -511,12 +511,12 @@ def _ci_limits_corr(self, corr0):
Difference between log-likelihood of corr0 and a pre-specified
value.
"""
return self.test_corr(corr0, nuis0=None, mu1_min=self.mu1_lb,
mu1_max=self.mu1_ub, mu2_min=self.mu2_lb,
mu2_max=self.mu2_ub,
var1_min=self.var1_lb, var1_max=self.var1_ub,
var2_min=self.var2_lb, var2_max=self.var2_ub,
print_weights=0)[0] - self.r0
return self.test_corr(corr0, nuis0=None, mu1_lb=self.mu1_lb,
mu1_ub=self.mu1_ub, mu2_lb=self.mu2_lb,
mu2_ub=self.mu2_ub,
var1_lb=self.var1_lb, var1_ub=self.var1_ub,
var2_lb=self.var2_lb, var2_ub=self.var2_ub,
print_weights=0, _is_ci=1)[0] - self.r0


class DescStatUV(_OptFuncts):
Expand Down Expand Up @@ -1390,10 +1390,11 @@ def mv_mean_contour(self, mu1_l, mu1_u, mu2_l, mu2_u, step1, step2,
ax.plot(self.endog[:, 0], self.endog[:, 1], 'bo')
return fig

def test_corr(self, corr0, nuis0=None, mu1_min=None,
mu1_max=None, mu2_min=None, mu2_max=None,
var1_min=None, var1_max=None,
var2_min=None, var2_max=None, print_weights=0):
def test_corr(self, corr0, nuis0=None, mu1_lb=None,
mu1_ub=None, mu2_lb=None, mu2_ub=None,
var1_lb=None, var1_ub=None,
var2_lb=None, var2_ub=None, print_weights=0,
_is_ci=False):
"""
Returns the -2 * log-likelihood ratio and p-value for the
correlation coefficient between 2 variables.
Expand Down Expand Up @@ -1434,9 +1435,7 @@ def test_corr(self, corr0, nuis0=None, mu1_min=None,
endog = self.endog
if endog.shape[1] != 2:
raise Exception('Correlation matrix not yet implemented')

self.corr0 = corr0

if nuis0 is not None:
start_nuisance = nuis0
else:
Expand All @@ -1445,50 +1444,45 @@ def test_corr(self, corr0, nuis0=None, mu1_min=None,
endog[:, 1].mean(),
endog[:, 1].var()])

if mu1_min is not None:
mu1_lb = mu1_min
else:
if _is_ci:
llr = optimize.fmin_l_bfgs_b(self._opt_correl, start_nuisance,
approx_grad=1,
bounds=[(mu1_lb, mu1_ub),
(var1_lb, var1_ub),
(mu2_lb, mu2_ub),
(var2_lb, var2_ub)])[1]
p_val = chi2.sf(llr, 1)
return llr, p_val

if mu1_lb is None:
mu1_lb = endog[:, 0].mean() - ((1.96 * \
np.sqrt((endog[:, 0].var()) / nobs)))

if mu1_max is not None:
mu1_ub = mu1_max
else:
if mu1_ub is None:
mu1_ub = endog[:, 0].mean() + (1.96 * \
np.sqrt((((endog[:, 0].var()) / nobs))))

if mu2_min is not None:
mu2_lb = mu2_min
else:
if mu2_lb is None:
mu2_lb = endog[:, 1].mean() - (1.96 * \
np.sqrt((((endog[:, 1].var()) / nobs))))

if mu2_max is not None:
mu2_ub = mu2_max
else:
if mu2_ub is None:
mu2_ub = endog[:, 1].mean() + (1.96 * \
np.sqrt((((endog[:, 1].var()) / nobs))))

if var1_min is not None:
var1_lb = var1_min
else:
if var1_lb is None:
var1_lb = (endog[:, 0].var() * (nobs - 1)) / \
chi2.ppf(.975, nobs)

if var1_max is not None:
var1_ub = var1_max
else:
if var1_ub is None:
var1_ub = (endog[:, 0].var() * (nobs - 1)) / \
chi2.ppf(.025, nobs)

if var2_min is not None:
var2_lb = var2_min
else:
if var2_lb is None:
var2_lb = (endog[:, 1].var() * (nobs - 1)) / \
chi2.ppf(.975, nobs)
if var2_max is not None:
var2_ub = var2_max
else:

if var2_ub is None:
var2_ub = (endog[:, 1].var() * (nobs - 1)) / \
chi2.ppf(.025, nobs)

Expand All @@ -1505,10 +1499,10 @@ def test_corr(self, corr0, nuis0=None, mu1_min=None,
return llr, p_val

def ci_corr(self, sig=.05, upper_bound=None, lower_bound=None,
mu1_min=None,
mu1_max=None, mu2_min=None, mu2_max=None,
var1_min=None, var1_max=None,
var2_min=None, var2_max=None):
mu1_lb=None,
mu1_ub=None, mu2_lb=None, mu2_ub=None,
var1_lb=None, var1_ub=None,
var2_lb=None, var2_ub=None):
"""
Returns the confidence intervals for the correlation coefficient.
Expand All @@ -1525,7 +1519,7 @@ def ci_corr(self, sig=.05, upper_bound=None, lower_bound=None,
Minimum value the lower condidence limit can be.
Default: 99% confidence assuming normality.
mu1_max through var2_max: float
mu1_ub through var2_ub: float
Limits of nuisance parameters
to maximize over. Default: 95% confidence limits assuming
normality
Expand Down Expand Up @@ -1557,50 +1551,50 @@ def ci_corr(self, sig=.05, upper_bound=None, lower_bound=None,
2.5 * (np.sqrt((1. - point_est ** 2.) / \
(nobs - 2.))))

if mu1_min is not None:
self.mu1_lb = mu1_min
if mu1_lb is not None:
self.mu1_lb = mu1_lb
else:
self.mu1_lb = endog[:, 0].mean() - np.sqrt((1.96 * \
(endog[:, 0].var()) / nobs))

if mu1_max is not None:
self.mu1_ub = mu1_max
if mu1_ub is not None:
self.mu1_ub = mu1_ub
else:
self.mu1_ub = endog[:, 0].mean() + (1.96 * \
np.sqrt(((endog[:, 0].var()) / nobs)))

if mu2_min is not None:
self.mu2_lb = mu2_min
if mu2_lb is not None:
self.mu2_lb = mu2_lb
else:
self.mu2_lb = endog[:, 1].mean() - (1.96 * \
np.sqrt(((endog[:, 1].var()) / nobs)))

if mu2_max is not None:
self.mu2_ub = mu2_max
if mu2_ub is not None:
self.mu2_ub = mu2_ub
else:
self.mu2_ub = endog[:, 1].mean() + (1.96 * \
np.sqrt(((endog[:, 1].var()) / nobs)))

if var1_min is not None:
self.var1_lb = var1_min
if var1_lb is not None:
self.var1_lb = var1_lb
else:
self.var1_lb = (endog[:, 0].var() * (nobs - 1)) / \
chi2.ppf(.975, nobs)

if var1_max is not None:
self.var1_ub = var1_max
if var1_ub is not None:
self.var1_ub = var1_ub
else:
self.var1_ub = (endog[:, 0].var() * (nobs - 1)) / \
chi2.ppf(.025, nobs)

if var2_min is not None:
self.var2_lb = var2_min
if var2_lb is not None:
self.var2_lb = var2_lb
else:
self.var2_lb = (endog[:, 1].var() * (nobs - 1)) / \
chi2.ppf(.975, nobs)

if var2_max is not None:
self.var2_ub = var2_max
if var2_ub is not None:
self.var2_ub = var2_ub
else:
self.var2_ub = (endog[:, 1].var() * (nobs - 1)) / \
chi2.ppf(.025, nobs)
Expand Down

0 comments on commit 5d081c9

Please sign in to comment.