Fixed a failing doctest. Reorganized _penalized_nnlf

wafo/spectrum/core.py

@@ -1429,7 +1429,9 @@ class SpecData1D(PlotData):
         >>> f = S.to_t_pdf(pdef='Tc', paramt=(0, 10, 51), speed=7)
         >>> Hm0 = 7
         >>> Tp = 11
-        >>> S = sm.Jonswap(4*pi/Tp,[Hm0 Tp])
+        >>> S = sm.Jonswap(4*pi/Tp,[Hm0, Tp])
+
+
         Sk = spec2spec(S,'k1d')
         L0 = spec2mom(S,1)
         paramu = [sqrt(L0)*[-4 4] 41]

wafo/stats/_distn_infrastructure.py

@@ -329,32 +329,32 @@ def nlogps(self, theta, x):
 
 
 def _penalized_nnlf(self, theta, x):
-        ''' Return negative loglikelihood function,
+    ''' Return negative loglikelihood function,
-        i.e., - sum (log pdf(x, theta), axis=0)
+    i.e., - sum (log pdf(x, theta), axis=0)
-           where theta are the parameters (including loc and scale)
+       where theta are the parameters (including loc and scale)
-        '''
+    '''
-        try:
+    try:
-            loc = theta[-2]
+        loc = theta[-2]
-            scale = theta[-1]
+        scale = theta[-1]
-            args = tuple(theta[:-2])
+        args = tuple(theta[:-2])
-        except IndexError:
+    except IndexError:
-            raise ValueError("Not enough input arguments.")
+        raise ValueError("Not enough input arguments.")
-        if not self._argcheck(*args) or scale <= 0:
+    if not self._argcheck(*args) or scale <= 0:
-            return inf
+        return inf
-        x = asarray((x-loc) / scale)
+    x = asarray((x-loc) / scale)
-
+    N = len(x)
-        loginf = log(_XMAX)
+    cond0 = (x < self.a) | (self.b < x)
-
+    Nbad = sum(cond0)
-        if np.isneginf(self.a).all() and np.isinf(self.b).all():
+    if Nbad > 0:
-            Nbad = 0
+        x = argsreduce(~cond0, x)[0]
-        else:
+    logpdf = self._logpdf(x, *args)
-            cond0 = (x < self.a) | (self.b < x)
+    finite_logpdf = np.isfinite(logpdf)
-            Nbad = sum(cond0)
+    Nbad += np.sum(~finite_logpdf, axis=0)
-            if Nbad > 0:
+    logscale = N * log(scale)
-                x = argsreduce(~cond0, x)[0]
+    if Nbad > 0:
-
+        penalty = Nbad * log(_XMAX) * 100
-        N = len(x)
+        return -np.sum(logpdf[finite_logpdf], axis=0) + penalty + logscale
-        return self._nnlf(x, *args) + N*log(scale) + Nbad * 100.0 * loginf
+    return -np.sum(logpdf, axis=0) + logscale
 
 
 def _reduce_func(self, args, options):