MAINT: undo filliben-related changes to the public scipy interface

scipy/stats/__init__.py

@@ -241,7 +241,6 @@
    mannwhitneyu
    tiecorrect
    rankdata
-   filliben
    ranksums
    wilcoxon
    kruskal

scipy/stats/morestats.py

@@ -317,6 +317,69 @@ def kstatvar(data, n=2):
         raise ValueError("Only n=1 or n=2 supported.")
 
 
+def _calc_uniform_order_statistic_medians(n):
+    """
+    Approximations of uniform order statistic medians.
+
+    Parameters
+    ----------
+    n : int
+        Sample size.
+
+    Returns
+    -------
+    v : 1d float array
+        Approximations of the order statistic medians.
+
+    References
+    ----------
+    .. [1] James J. Filliben, "The Probability Plot Correlation Coefficient
+           Test for Normality", Technometrics, Vol. 17, pp. 111-117, 1975.
+
+    Examples
+    --------
+    Order statistics of the uniform distribution on the unit interval
+    are marginally distributed according to beta distributions.
+    The expectations of these order statistic are evenly spaced across
+    the interval, but the distributions are skewed in a way that
+    pushes the medians slightly towards the endpoints of the unit interval:
+
+    >>> n = 4
+    >>> k = np.arange(1, n+1)
+    >>> from scipy.stats import beta
+    >>> a = k
+    >>> b = n-k+1
+    >>> beta.mean(a, b)
+    array([ 0.2,  0.4,  0.6,  0.8])
+    >>> beta.median(a, b)
+    array([ 0.15910358,  0.38572757,  0.61427243,  0.84089642])
+
+    The Filliben approximation uses the exact medians of the smallest
+    and greatest order statistics, and the remaining medians are approximated
+    by points spread evenly across a sub-interval of the unit interval:
+
+    >>> from scipy.morestats import _calc_uniform_order_statistic_medians
+    >>> _calc_uniform_order_statistic_medians(n)
+    array([ 0.15910358,  0.38545246,  0.61454754,  0.84089642])
+
+    This plot shows the skewed distributions of the order statistics
+    of a sample of size four from a uniform distribution on the unit interval:
+
+    >>> import matplotlib.pyplot as plt
+    >>> x = np.linspace(0.0, 1.0, num=50, endpoint=True)
+    >>> pdfs = [beta.pdf(x, a[i], b[i]) for i in range(n)]
+    >>> plt.figure()
+    >>> plt.plot(x, pdfs[0], x, pdfs[1], x, pdfs[2], x, pdfs[3])
+
+    """
+    v = np.zeros(n, dtype=np.float64)
+    v[-1] = 0.5**(1.0 / n)
+    v[0] = 1 - v[-1]
+    i = np.arange(2, n)
+    v[1:-1] = (i - 0.3175) / (n + 0.365)
+    return v
+
+
 def _parse_dist_kw(dist, enforce_subclass=True):
     """Parse `dist` keyword.
 
@@ -486,7 +549,7 @@ def probplot(x, sparams=(), dist='norm', fit=True, plot=None):
         else:
             return x, x
 
-    osm_uniform = stats.filliben(len(x))
+    osm_uniform = _calc_uniform_order_statistic_medians(len(x))
     dist = _parse_dist_kw(dist, enforce_subclass=False)
     if sparams is None:
         sparams = ()
@@ -598,7 +661,7 @@ def ppcc_max(x, brack=(0.0, 1.0), dist='tukeylambda'):
 
     """
     dist = _parse_dist_kw(dist)
-    osm_uniform = stats.filliben(len(x))
+    osm_uniform = _calc_uniform_order_statistic_medians(len(x))
     osr = sort(x)
 
     # this function computes the x-axis values of the probability plot
@@ -1009,7 +1072,7 @@ def boxcox_normmax(x, brack=(-2.0, 2.0), method='pearsonr'):
     """
 
     def _pearsonr(x, brack):
-        osm_uniform = stats.filliben(len(x))
+        osm_uniform = _calc_uniform_order_statistic_medians(len(x))
         xvals = distributions.norm.ppf(osm_uniform)
 
         def _eval_pearsonr(lmbda, xvals, samps):

scipy/stats/stats.py

@@ -157,7 +157,6 @@
    ss
    square_of_sums
    rankdata
-   filliben
 
 References
 ----------
@@ -190,7 +189,7 @@
 __all__ = ['find_repeats', 'gmean', 'hmean', 'mode', 'tmean', 'tvar',
            'tmin', 'tmax', 'tstd', 'tsem', 'moment', 'variation',
            'skew', 'kurtosis', 'describe', 'skewtest', 'kurtosistest',
-           'normaltest', 'jarque_bera', 'filliben', 'itemfreq',
+           'normaltest', 'jarque_bera', 'itemfreq',
            'scoreatpercentile', 'percentileofscore', 'histogram',
            'histogram2', 'cumfreq', 'relfreq', 'obrientransform',
            'signaltonoise', 'sem', 'zmap', 'zscore', 'threshold',
@@ -1559,69 +1558,6 @@ def jarque_bera(x):
     return jb_value, p
 
 
-def filliben(n):
-    """
-    Approximations of uniform order statistic medians.
-
-    Parameters
-    ----------
-    n : int
-        Sample size.
-
-    Returns
-    -------
-    v : 1d float array
-        Approximations of the order statistic medians.
-
-    References
-    ----------
-    .. [1] James J. Filliben, "The Probability Plot Correlation Coefficient
-           Test for Normality", Technometrics, Vol. 17, pp. 111-117, 1975.
-
-    Examples
-    --------
-    Order statistics of the uniform distribution on the unit interval
-    are marginally distributed according to beta distributions.
-    The expectations of these order statistic are evenly spaced across
-    the interval, but the distributions are skewed in a way that
-    pushes the medians slightly towards the endpoints of the unit interval:
-
-    >>> n = 4
-    >>> k = np.arange(1, n+1)
-    >>> from scipy.stats import beta
-    >>> a = k
-    >>> b = n-k+1
-    >>> beta.mean(a, b)
-    array([ 0.2,  0.4,  0.6,  0.8])
-    >>> beta.median(a, b)
-    array([ 0.15910358,  0.38572757,  0.61427243,  0.84089642])
-
-    The Filliben approximation uses the exact medians of the smallest
-    and greatest order statistics, and the remaining medians are approximated
-    by points spread evenly across a sub-interval of the unit interval:
-
-    >>> from scipy.stats import filliben
-    >>> filliben(n)
-    array([ 0.15910358,  0.38545246,  0.61454754,  0.84089642])
-
-    This plot shows the skewed distributions of the order statistics
-    of a sample of size four from a uniform distribution on the unit interval:
-
-    >>> import matplotlib.pyplot as plt
-    >>> x = np.linspace(0.0, 1.0, num=50, endpoint=True)
-    >>> pdfs = [beta.pdf(x, a[i], b[i]) for i in range(n)]
-    >>> plt.figure()
-    >>> plt.plot(x, pdfs[0], x, pdfs[1], x, pdfs[2], x, pdfs[3])
-
-    """
-    v = np.zeros(n, dtype=np.float64)
-    v[-1] = 0.5**(1.0 / n)
-    v[0] = 1 - v[-1]
-    i = np.arange(2, n)
-    v[1:-1] = (i - 0.3175) / (n + 0.365)
-    return v
-
-
 #####################################
 #        FREQUENCY FUNCTIONS        #
 #####################################