def isf(self, q, *args, **kwds):
"""
Inverse survival function at q of the given RV.
Parameters
----------
q : array-like
upper tail probability
arg1, arg2, arg3,... : array-like
The shape parameter(s) for the distribution (see docstring of the
instance object for more information)
loc : array-like, optional
location parameter (default=0)
scale : array-like, optional
scale parameter (default=1)
Returns
-------
x : array-like
quantile corresponding to the upper tail probability q.
"""
(loc, scale) = map(kwds.get, ['loc', 'scale'])
(args, loc, scale) = self._fix_loc_scale(args, loc, scale)
(q, loc, scale, shape) = map(arr, (q, loc, scale, args[0]))
q = np.where(scale > 0, q, 1. - q)
#
isgamma = (shape > 0) & (scale != 0)
isnorm = (shape == 0) & (scale > 0)
ispe3 = (isgamma | isnorm)
indom = (q > 0) & (q < 1)
islarge = (q == 1) & ispe3
valid = ispe3 & indom
output = valarray(np.shape(valid), value=self.b)
#place(output,(1-cond0)*(cond1==cond1), self.badvalue)
np.place(output, (1 - ispe3) * (indom == indom) + (1 - indom) * (q != 0.0),
self.badvalue)
np.place(output, islarge, self.a)
if np.any(valid): #call only if at least 1 entry
goodargs = argsreduce(valid, *((q,) + args + (scale, loc))) #PB replace 1-q by q
(q, shape, scale, loc) = argsreduce(valid, *((q, shape, scale, loc)))
np.place(output, (valid & isgamma), self._isf(q, shape) * scale + loc)
np.place(output, (valid & isnorm), -dist._norm_ppf(q))
if output.ndim == 0:
return output[()]
return output
def _ppf(self, q, k):
u = dist._norm_ppf(q)
return where(k == 0, u, (1. - exp(-k * u)) / k)
