Update the unit tests and fix some bugs

Author: LI-Yixuan

dabest/_classes.py

@@ -368,7 +368,7 @@ def __repr__(self):
                     comparisons.append("{} minus {}".format(test_name, control_name))
 
         if self.__delta2:
-            comparison.append("{} minus {}".format(self.__experiment_label[1], self.__experiment_label[0]))
+            comparisons.append("{} minus {} (only for mean difference)".format(self.__experiment_label[1], self.__experiment_label[0]))
 
         for j, g in enumerate(comparisons):
             out.append("{}. {}".format(j+1, g))
@@ -782,6 +782,9 @@ def __init__(self, control, test, effect_size,
             `random_seed` is used to seed the random number generator during
             bootstrap resampling. This ensures that the confidence intervals
             reported are replicable.
+        delta2 : boolean, default False
+            Indicate if the control and test data are boostrap deltas that can be
+            used to calculate delta-delta.
 
 
         Returns
@@ -1145,7 +1148,7 @@ def __repr__(self, show_resample_count=True, define_pval=True, sigfig=3):
                       "paired_status": PAIRED_STATUS[str(self.__is_paired)]}
 
         if self.__delta2:
-            out1 = "The delta-delta {es}"
+            out1 = "The delta-delta "
         else:
             out1 = "The {paired_status} {es} {rm_status}".format(**first_line)
 
@@ -1594,8 +1597,12 @@ def __pre_calc(self):
                 r_dict["test_N"]    = int(len(test))
                 out.append(r_dict)
                 if j == len(idx)-1 and ix == len(current_tuple)-2:
-                    resamp_count = True
-                    def_pval     = True
+                    if not self.__delta2:
+                        resamp_count = True
+                        def_pval     = True
+                    else:
+                        resamp_count = False
+                        def_pval     = False
                 else:
                     resamp_count = False
                     def_pval     = False
@@ -1612,25 +1619,30 @@ def __pre_calc(self):
             delta = TwoGroupsEffectSize(out[0]["bootstraps"], 
                                     out[1]["bootstraps"],
                                     self.__effect_size,
-                                    True,
+                                    "baseline",
                                     self.__ci,
                                     self.__resamples,
                                     self.__permutation_count,
                                     self.__random_seed,
                                     self.__delta2
                                     )
+
             r_dict = delta.to_dict()
             r_dict["control"]   = self.__experiment_label[0]
             r_dict["test"]      = self.__experiment_label[1]
             r_dict["control_N"] = self.__resamples
             r_dict["test_N"]    = self.__resamples
             out.append(r_dict)
+            resamp_count = True
+            def_pval     = True
+            text_repr = delta.__repr__(show_resample_count=resamp_count,
+                                            define_pval=def_pval)
             to_replace = "between {} and {} is".format(self.__experiment_label[0], self.__experiment_label[1])
             text_repr = text_repr.replace("is", to_replace, 1)
             reprs.append(text_repr)
         else:
-            err0 = 'The calculation of delta-delta is not supported for {}.'.format(self.__effect_size)
-            raise ValueError(err0)
+            self.__delta2 = False
+
         varname = get_varname(self.__dabest_obj)
         lastline = "To get the results of all valid statistical tests, " +\
         "use `{}.{}.statistical_tests`".format(varname, self.__effect_size)

dabest/plotter.py

@@ -59,16 +59,19 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
     ytick_color = plt.rcParams["ytick.color"]
     axes_facecolor = plt.rcParams['axes.facecolor']
 
-    dabest_obj = EffectSizeDataFrame.dabest_obj
-    plot_data  = EffectSizeDataFrame._plot_data
-    xvar       = EffectSizeDataFrame.xvar
-    yvar       = EffectSizeDataFrame.yvar
-    is_paired  = EffectSizeDataFrame.is_paired
-    delta2     = EffectSizeDataFrame.delta2
+    dabest_obj  = EffectSizeDataFrame.dabest_obj
+    plot_data   = EffectSizeDataFrame._plot_data
+    xvar        = EffectSizeDataFrame.xvar
+    yvar        = EffectSizeDataFrame.yvar
+    is_paired   = EffectSizeDataFrame.is_paired
+    delta2      = EffectSizeDataFrame.delta2
+    effect_size = EffectSizeDataFrame.effect_size
 
     all_plot_groups = dabest_obj._all_plot_groups
     idx             = dabest_obj.idx
 
+    if effect_size != "mean_diff":
+        delta2 = False
 
 
 
@@ -93,6 +96,7 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
         show_pairs = plot_kwargs["show_pairs"]
 
 
+
     # Set default kwargs first, then merge with user-dictated ones.
     default_swarmplot_kwargs = {'size': plot_kwargs["raw_marker_size"]}
     if plot_kwargs["swarmplot_kwargs"] is None:
@@ -102,6 +106,7 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
                                            plot_kwargs["swarmplot_kwargs"])
 
 
+
     # Violinplot kwargs.
     default_violinplot_kwargs = {'widths':0.5, 'vert':True,
                                'showextrema':False, 'showmedians':False}
@@ -112,6 +117,7 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
                                             plot_kwargs["violinplot_kwargs"])
 
 
+
     # slopegraph kwargs.
     default_slopegraph_kwargs = {'lw':1, 'alpha':0.5}
     if plot_kwargs["slopegraph_kwargs"] is None:
@@ -121,6 +127,7 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
                                             plot_kwargs["slopegraph_kwargs"])
 
 
+
     # Zero reference-line kwargs.
     default_reflines_kwargs = {'linestyle':'solid', 'linewidth':0.75,
                                 'zorder': 2,
@@ -218,6 +225,7 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
     plot_palette_contrast = dict(zip(names, contrast_colors))
 
 
+
     # Infer the figsize.
     fig_size   = plot_kwargs["fig_size"]
     if fig_size is None:
@@ -322,6 +330,7 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
     contrast_axes.set_frame_on(False)
     if delta2:
         delta_axes.set_frame_on(False)
+
     redraw_axes_kwargs = {'colors'     : ytick_color,
                           'facecolors' : ytick_color,
                           'lw'      : 1,
@@ -460,8 +469,6 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
         ticks_to_plot = np.arange(1, len(temp_all_plot_groups), 2).tolist() 
         ticks_to_skip_contrast = np.cumsum([(len(t)-1)*2 for t in idx])[:-1].tolist()
         ticks_to_skip_contrast.insert(0, 0)
-        if delta2:
-            ticks_to_skip_contrast.append(max(ticks_to_skip_contrast)+2)
     else:
         ticks_to_skip   = np.cumsum([len(t) for t in idx])[:-1].tolist()
         ticks_to_skip.insert(0, 0)
@@ -575,6 +582,8 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
     if float_contrast is True:
         contrast_axes.set_xlim(0.5, 1.5)
     elif delta2:
+        # Increase the xlim of raw data by 2; set xlim and label xlim 
+        # of the delta-delta plot.
         contrast_axes.set_xlim(rawdata_axes.get_xlim())
         temp = rawdata_axes.get_xlim()
         rawdata_axes.set_xlim(temp[0], temp[1]+2)
@@ -769,7 +778,8 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
             contrast_ylim_high, contrast_ylim_low = contrast_axes_ylim
         if contrast_ylim_low < 0 < contrast_ylim_high:
             contrast_axes.axhline(y=0, **reflines_kwargs)
-            delta_axes.axhline(y=0, xmin = 0.25, xmax = 1.5, **reflines_kwargs)
+            if delta2:
+                delta_axes.axhline(y=0, xmin = 0.25, xmax = 1.5, **reflines_kwargs)
 
 
         if delta2 and plot_kwargs['delta_ylim'] is not None:
@@ -789,7 +799,7 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
                 delta_axes.set_ylim(low, high)
             else:
                 delta_axes.set_ylim(custom_delta_ylim)
-        if delta2:
+        if delta2 and plot_kwargs['delta_ylim'] is None and plot_kwargs['contrast_ylim'] is None:
             ylim_contrast = contrast_axes.get_ylim()
             ylim_delta = delta_axes.get_ylim()
             ylim=(min(ylim_contrast[0], ylim_delta[0]), max(ylim_contrast[1], ylim_delta[1]))
@@ -814,8 +824,6 @@ def EffectSizeDataFramePlotter(EffectSizeDataFrame, **plot_kwargs):
                 del redraw_axes_kwargs['y']
 
             temp_length = [(len(i)-1)*2-1 for i in idx]
-            if delta2:
-                temp_length.append(1)
             rightend_ticks_contrast = np.array(temp_length) + np.array(ticks_to_skip_contrast)
             for ax in [contrast_axes]:
                 sns.despine(ax=ax, bottom=True)

dabest/tests/test_01_effsizes_pvals.py

@@ -55,7 +55,10 @@
 
 # kwargs for Dabest class init.
 dabest_default_kwargs = dict(x=None, y=None, ci=95, 
-                            resamples=5000, random_seed=12345)
+                            resamples=5000, random_seed=12345,
+                            proportional=False, delta2=False, experiment=None, 
+                            experiment_label=None, x1_level=None
+                            )
 
 
 
@@ -79,7 +82,7 @@ def test_mean_diff_paired():
     from numpy import mean as npmean
     mean_diff = effsize.func_difference(paired_wellbeing.pre,
                                         paired_wellbeing.post,
-                                        npmean, is_paired=True)
+                                        npmean, is_paired="baseline")
     assert mean_diff == pytest.approx(4.10)
 
 
@@ -88,7 +91,7 @@ def test_median_diff_paired():
     from numpy import median as npmedian
     median_diff = effsize.func_difference(paired_wellbeing.pre,
                                           paired_wellbeing.post,
-                                          npmedian, is_paired=True)
+                                          npmedian, is_paired="baseline")
     assert median_diff == pytest.approx(4.5)
 
 
@@ -112,15 +115,15 @@ def test_hedges_g_unpaired():
 def test_cohens_d_paired():
     import numpy as np
     cohens_d = effsize.cohens_d(paired_wellbeing.pre, paired_wellbeing.post,
-                                is_paired=True)
+                                is_paired="baseline")
     assert np.round(cohens_d, 2) == pytest.approx(0.34)
 
 
 
 def test_hedges_g_paired():
     import numpy as np
     hedges_g = effsize.hedges_g(paired_wellbeing.pre, paired_wellbeing.post,
-                                is_paired=True)
+                                is_paired="baseline")
     assert np.round(hedges_g, 2) == pytest.approx(0.33)
 
 
@@ -155,7 +158,7 @@ def test_paired_stats():
     before = paired_wellbeing.pre
     after = paired_wellbeing.post
 
-    paired_es = TwoGroupsEffectSize(before, after, "mean_diff", is_paired=True)
+    paired_es = TwoGroupsEffectSize(before, after, "mean_diff", is_paired="baseline")
 
     p1 = sp.stats.ttest_rel(before, after, nan_policy='omit').pvalue
     assert paired_es.pvalue_paired_students_t == pytest.approx(p1)
@@ -197,7 +200,7 @@ def test_paired_permutation_test():
     perm_test = PermutationTest(paired_wellbeing.pre, 
                                 paired_wellbeing.post, 
                                 effect_size="mean_diff", 
-                                is_paired=True)
+                                is_paired="baseline")
     assert perm_test.pvalue == pytest.approx(0.0124)
 
 
@@ -223,7 +226,7 @@ def test_lqrt_unpaired():
 
 def test_lqrt_paired():
     paired_dabest = Dabest(paired_wellbeing, idx=("pre", "post"),
-                           paired=True, id_col="ID",
+                           paired="baseline", id_col="ID",
                            **dabest_default_kwargs)
     lqrt_result = paired_dabest.mean_diff.lqrt
 

dabest/tests/test_03_plotting.py

@@ -25,7 +25,7 @@
 two_groups_unpaired = load(df, idx=("Control 1", "Test 1"))
 
 two_groups_paired   = load(df, idx=("Control 1", "Test 1"),
-                           paired=True, id_col="ID")
+                           paired="baseline", id_col="ID")
 
 multi_2group = load(df, idx=(("Control 1", "Test 1",),
                              ("Control 2", "Test 2"))
@@ -34,7 +34,7 @@
 multi_2group_paired = load(df,
                             idx=(("Control 1", "Test 1"),
                                  ("Control 2", "Test 2")),
-                            paired=True, id_col="ID")
+                            paired="baseline", id_col="ID")
 
 shared_control = load(df, idx=("Control 1", "Test 1",
                                 "Test 2", "Test 3",
@@ -137,7 +137,7 @@ def test_11_inset_plots():
     iris_dabest3 = load(data=iris_melt[iris_melt.species=="setosa"],
                         x="metric", y="value",
                         idx=("sepal_length", "sepal_width"),
-                        paired=True, id_col="index")
+                        paired="baseline", id_col="index")