format code

Author: qbarthelemy

pypermut/__init__.py

@@ -1,3 +1,3 @@
 """ PyPermut """
 
-__version__ = "0.1.0"
+__version__ = "0.2.0"

pypermut/core.py

@@ -62,7 +62,8 @@ def permute_measurements(Y, x,
     perms, n_perms, with_replacement = helpers._check_permutations(
         n_perms_requested=n_permutations,
         n_perms_max=n_perms_max,
-        with_replacement=with_replacement)
+        with_replacement=with_replacement,
+    )
 
     # loop on permutations to sample the null distribution
     null_dist = np.empty(n_perms, dtype=float)
@@ -73,7 +74,8 @@ def permute_measurements(Y, x,
         else:
             permuted_indices = get_permutation_measurements(
                 n_meas,
-                perms[i_perm])
+                perms[i_perm],
+            )
         xperm = x[permuted_indices]  # permute x along measurements
 
         stat = stat_func(np.c_[xperm, Y])
@@ -144,7 +146,8 @@ def permute_paired_samples(X, Y,
     perms, n_perms, with_replacement = helpers._check_permutations(
         n_perms_requested=n_permutations,
         n_perms_max=n_perms_max,
-        with_replacement=with_replacement)
+        with_replacement=with_replacement,
+    )
 
     # loop on permutations to sample the null distribution
     null_dist = np.empty(n_perms, dtype=float)
@@ -158,7 +161,8 @@ def permute_paired_samples(X, Y,
         else:
             perm_coeffs = get_permutation_2_paired_samples(
                 n_meas,
-                perms[i_perm])
+                perms[i_perm],
+            )
         # permute each pair: apply random coeffs -1 or +1 on differences
         Dperm = np.tile(perm_coeffs, D.shape[1]) * D
 
@@ -233,7 +237,8 @@ def permute_unpaired_samples(args,
     perms, n_perms, with_replacement = helpers._check_permutations(
         n_perms_requested=n_permutations,
         n_perms_max=n_perms_max,
-        with_replacement=with_replacement)
+        with_replacement=with_replacement,
+    )
     if not with_replacement:
         if len(args) != 2:
             raise ValueError('Without replacement strategy is available only '
@@ -249,7 +254,8 @@ def permute_unpaired_samples(args,
         else:
             permuted_indices = get_permutation_unpaired_samples(
                 list_meas,
-                list(combs[perms[i_perm]]))
+                list(combs[perms[i_perm]]),
+            )
         Cperm = C[permuted_indices]  # permute measurements between samples
 
         stat = stat_func(Cperm, list_meas)
@@ -406,8 +412,10 @@ def get_permutation_2_paired_samples(n_meas, perm_number):
         raise IndexError('Permutation number {} is out of range [0 ... {}].'
                          .format(perm_number, 2**n_meas))
     # transform number into binary
-    bin_coeffs = np.fromiter(np.binary_repr(perm_number, width=n_meas),
-                             dtype=int)
+    bin_coeffs = np.fromiter(
+        np.binary_repr(perm_number, width=n_meas),
+        dtype=int,
+    )
     # transform binaries 0/1 into coefficients 1/-1
     perm_coeffs = 1 - 2 * np.array(bin_coeffs)[:, np.newaxis]
 

pypermut/helpers.py

@@ -64,7 +64,8 @@ def _check_permutations(n_perms_requested, n_perms_max, with_replacement):
             perms = np.random.choice(
                 np.arange(0, n_perms_max),
                 size=n_perms,
-                replace=False)
+                replace=False,
+            )
         else:
             # => permutation test, using bootstrap with replacement
             perms = []

pypermut/stats.py

@@ -6,7 +6,7 @@
 
 import warnings
 import numpy as np
-import scipy.stats as stats
+from scipy.stats import percentileofscore
 from . import helpers
 from . import core
 from . import mstats
@@ -129,7 +129,8 @@ def permutation_corr(Y, *,
         with_replacement,
         stat_func=corr_func,
         var_func=np.max,
-        side=side)
+        side=side,
+    )
 
     # compute the real R statistics
     Rstats = Rstats_ = corr_func(np.c_[x, Y])
@@ -138,8 +139,9 @@ def permutation_corr(Y, *,
     # compare them to the Rmax distribution with a right-sided test,
     # because significance is obtained for high R stats
     pvals = np.array([
-        (100 - stats.percentileofscore(Rmax, R, kind='strict')) / 100
-        for R in Rstats_])
+        (100 - percentileofscore(Rmax, R, kind='strict')) / 100
+        for R in Rstats_
+    ])
 
     if return_dist:
         return Rstats, pvals, Rmax
@@ -223,7 +225,8 @@ def permutation_ttest_rel(X, Y, *,
         with_replacement=with_replacement,
         stat_func=mstats.studentt_rel,
         var_func=np.max,
-        side=side)
+        side=side,
+    )
 
     # compute the real t statistics
     tstats = tstats_ = mstats.studentt_rel(X - Y)
@@ -232,8 +235,9 @@ def permutation_ttest_rel(X, Y, *,
     # compare them to the tmax distribution with a right-sided test,
     # because significance is obtained for high t stats
     pvals = np.array([
-        (100 - stats.percentileofscore(tmax, t, kind='strict')) / 100
-        for t in tstats_])
+        (100 - percentileofscore(tmax, t, kind='strict')) / 100
+        for t in tstats_
+    ])
 
     if return_dist:
         return tstats, pvals, tmax
@@ -320,18 +324,22 @@ def permutation_ttest_ind(X, Y, *,
         with_replacement=with_replacement,
         stat_func=stat_func,
         var_func=np.max,
-        side=side)
+        side=side,
+    )
 
     # compute the real t statistics
-    tstats = tstats_ = stat_func(np.concatenate((X, Y), axis=0),
-                                 [X.shape[0], Y.shape[0]])
+    tstats = tstats_ = stat_func(
+        np.concatenate((X, Y), axis=0),
+        [X.shape[0], Y.shape[0]],
+    )
     if side == 'two':
         tstats_ = np.abs(tstats_)
     # compare it to the tmax distribution with a right-sided test,
     # because significance is obtained for high t stats
     pvals = np.array([
-        (100 - stats.percentileofscore(tmax, t, kind='strict')) / 100
-        for t in tstats_])
+        (100 - percentileofscore(tmax, t, kind='strict')) / 100
+        for t in tstats_
+    ])
 
     if return_dist:
         return tstats, pvals, tmax
@@ -407,14 +415,16 @@ def permutation_wilcoxon(X, Y, *,
         stat_func=mstats.wilcoxon,
         var_func=np.min,
         side='one',
-        zero_method=zero_method)
+        zero_method=zero_method,
+    )
 
     # compute the real T statistics
     Tstats = mstats.wilcoxon(X - Y, zero_method)
     # compare it to the Tmin distribution with a left-sided test,
     # because significance is obtained for low T stats
     pvals = np.array([
-        stats.percentileofscore(Tmin, T, kind='weak') / 100 for T in Tstats])
+        percentileofscore(Tmin, T, kind='weak') / 100 for T in Tstats
+    ])
 
     if return_dist:
         return Tstats, pvals, Tmin
@@ -482,15 +492,19 @@ def permutation_mannwhitneyu(X, Y, *,
         with_replacement=with_replacement,
         stat_func=mstats.mannwhitneyu,
         var_func=np.min,
-        side='one')
+        side='one',
+    )
 
     # compute the real U statistics
-    Ustats = mstats.mannwhitneyu(np.concatenate((X, Y), axis=0),
-                                 [X.shape[0], Y.shape[0]])
+    Ustats = mstats.mannwhitneyu(
+        np.concatenate((X, Y), axis=0),
+        [X.shape[0], Y.shape[0]],
+    )
     # compare it to the Umin distribution with a left-sided test,
     # because significance is obtained for low U stats
     pvals = np.array([
-        stats.percentileofscore(Umin, U, kind='weak') / 100 for U in Ustats])
+        percentileofscore(Umin, U, kind='weak') / 100 for U in Ustats
+    ])
 
     if return_dist:
         return Ustats, pvals, Umin
@@ -550,7 +564,8 @@ def permutation_f_oneway(*args, n=10000, return_dist=False):
         with_replacement=True,
         stat_func=mstats.f_oneway,
         var_func=np.max,
-        side='one')
+        side='one',
+    )
 
     # number of measurements for each sample / group
     list_meas = np.asarray(list(map(len, args)))
@@ -559,8 +574,8 @@ def permutation_f_oneway(*args, n=10000, return_dist=False):
     # compare it to the Fmax distribution with a right-sided test,
     # because significance is obtained for high F stats
     pvals = np.array([
-        (100 - stats.percentileofscore(Fmax, F, kind='strict')) / 100
-        for F in Fstats])
+        (100 - percentileofscore(Fmax, F, kind='strict')) / 100 for F in Fstats
+    ])
 
     if return_dist:
         return Fstats, pvals, Fmax
@@ -621,16 +636,17 @@ def permutation_kruskal(*args, n=10000, return_dist=False):
         with_replacement=True,
         stat_func=mstats.kruskal,
         var_func=np.max,
-        side='one')
+        side='one',
+    )
 
     list_meas = np.asarray(list(map(len, args)))
     # compute the real H statistics
     Hstats = mstats.kruskal(np.concatenate(args, axis=0), list_meas)
     # compare it to the Hmax distribution with a right-sided test,
     # because significance is obtained for high H stats
     pvals = np.array([
-        (100 - stats.percentileofscore(Hmax, H, kind='strict')) / 100
-        for H in Hstats])
+        (100 - percentileofscore(Hmax, H, kind='strict')) / 100 for H in Hstats
+    ])
 
     if return_dist:
         return Hstats, pvals, Hmax
@@ -714,8 +730,9 @@ def permutation_friedmanchisquare(*args, n=10000, return_dist=False):
     # compare it to the chimax distribution with a right-sided test,
     # because significance is obtained for high chi2 stats
     pvals = np.array([
-       (100 - stats.percentileofscore(chi2max, chi2, kind='strict')) / 100
-       for chi2 in chi2stats])
+        (100 - percentileofscore(chi2max, chi2, kind='strict')) / 100
+        for chi2 in chi2stats
+    ])
 
     if return_dist:
         return chi2stats, pvals, chi2max

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "pypermut"
-version = "0.1.0"
+version = "0.2.0"
 description = "PyPermut is a Python package implementing permutation tests"
 authors = ["Quentin Barthelemy", "Anonymous", "Louis Mayaud"]
 license = "BSD-3-Clause"