Clean up code to make sure we can recreate figures

activ/clustering/viz.py

@@ -570,10 +570,6 @@ def plot_real_foc_results(path, ax=None, max1d_cutoff=False, ci=None, n_sigma=1,
     ax = ax or plt.gca()
     real_noc, real_foc, real_accuracy, real_chance = read_clustering_results(path)
 
-    real_data_est = get_real_noc(np.arange(2, 51), real_foc,
-                                 smooth=True, use_median=False,
-                                 spread_asm=True, spread_foc=True)
-
 
     x_noc, lower, med, upper = flatten_summarize(np.arange(2, 51),
                                                  real_foc, smooth=True,
@@ -585,7 +581,7 @@ def plot_real_foc_results(path, ax=None, max1d_cutoff=False, ci=None, n_sigma=1,
         noc_idx = get_noc_max1d_smooth(np.arange(2, 51), real_foc)
     else:
         noc_idx = get_noc(np.arange(2, 51), real_foc, plot_fit=True, plot_asm=False, fit_summary=True,
-                          ttest_cutoff=False, iqr=False, ax=ax, use_median=False, spread_asm=True, spread_foc=True,
+                          ttest_cutoff=False, iqr=True, ax=ax, use_median=False, spread_asm=True, spread_foc=True,
                           ci=ci, n_sigma=n_sigma)
 
     est = np.arange(2, 51)[noc_idx]
@@ -668,7 +664,7 @@ def entropy_across_clusterings(variable, labels):
     return np.array(ret)
 
 
-def plot_entropy_across_clusters(features, emb, ax=None, cmap='Greys'):
+def plot_entropy_across_clusters(features, emb, ax=None, cmap='Greys', fontsize='x-large'):
     """
     Plot average feature entropy as a function of the number of clusters
     for every feature in *features*
@@ -698,20 +694,20 @@ def plot_entropy_across_clusters(features, emb, ax=None, cmap='Greys'):
     order = np.argsort(init)[::-1]
     for var_i in order:
         ax.plot(x, y[var_i]/init[var_i], c=cmap(init[var_i]))
-    ax.set_ylabel("Percent of initial entropy", fontsize='x-large')
-    ax.set_xlabel("Number of clusters", fontsize='x-large')
+    ax.set_ylabel("Percent of initial entropy", fontsize=fontsize)
+    ax.set_xlabel("Number of clusters", fontsize=fontsize)
 
     yticks = np.array([0.2, 0.4, 0.6, 0.8, 1.0])
 
     ax.set_yticks(yticks)
-    ax.set_yticklabels((yticks*100).astype(int), fontsize='x-large')
+    ax.set_yticklabels((yticks*100).astype(int), fontsize=fontsize)
 
     xticks = np.arange(6)*10
     ax.set_xticks(xticks)
-    ax.set_xticklabels(xticks, fontsize='x-large')
+    ax.set_xticklabels(xticks, fontsize=fontsize)
 
     cbar = ax.figure.colorbar(cm.ScalarMappable(norm=mpc.Normalize(vmin=np.min(init), vmax=np.max(init)), cmap=cmap), ax=ax)
-    cbar.set_label("Initial entropy", rotation=-90, va="bottom", fontsize='x-large')
+    cbar.set_label("Initial entropy", rotation=-90, va="bottom", fontsize=fontsize)
     cbar_yticks = [1, 2, 3, 4]
 
-    cbar.ax.set_yticklabels(cbar_yticks, fontsize='x-large')
+    cbar.ax.set_yticklabels(cbar_yticks, fontsize=fontsize)

activ/nmf/viz.py

@@ -300,6 +300,13 @@ def plot_weights(weights, colors=None, factor_order=None, ax=None, labels=None,
             ax_bar.spines['right'].set_visible(False)
             ax_bar.spines['top'].set_visible(False)
             ax_bar.set_ylabel('Fraction of\npatients', fontsize=fontsize)
+
+            yticks = np.array([100, 200])
+            ax_bar2 = ax_bar.twinx()
+            ax_bar2.set_yticks(yticks / weights.shape[0], labels=yticks, fontsize=labelsize)
+            ax_bar2.set_ylabel('Number of\npatients', fontsize=fontsize)
+            ax_bar2.spines['top'].set_visible(False)
+            ax_bar.spines['left'].set_visible(False)
             ax = ax[1]
         else:
             ax = ax[0]

activ/paper.py

@@ -0,0 +1,77 @@
+#import matplotlib.pyplot as plt
+import numpy as np
+from matplotlib.patches import Ellipse
+import matplotlib.transforms as transforms
+
+
+def draw_ellipse(x, y, width, height, ax=None, color='blue', linewidth=1, alpha=0.1):
+
+    #ax = ax or plt.gca()
+    face = Ellipse((x, y),
+        width=width,
+        height=height,
+        facecolor=color,
+        angle=45,
+        alpha=alpha)
+    edge = Ellipse((x, y),
+        width=width,
+        height=height,
+        angle=45,
+        edgecolor=color,
+        facecolor='none',
+        linewidth=linewidth)
+    ax.add_patch(face)
+    ax.add_patch(edge)
+
+
+def get_cuts(left, right, num_circles):
+    cuts = np.linspace(left, right, num_circles*2 + 1)
+    center_idx = np.arange(num_circles*2 + 1)[1:-1:2]
+    return cuts, center_idx
+
+
+def add_circles(left, right, num_circles, color, num_levels=1, current_level=0, minor_d_frac=1.0, ax=None):
+    if current_level >= num_levels:
+        return
+    #ax = ax or plt.gca()
+    cuts, center_idx = get_cuts(left, right, num_circles)
+    line_length = right-left
+    d = np.sqrt(2) * line_length / num_circles
+    for i, c_i in enumerate(center_idx):
+        center = cuts[c_i]
+        col = color
+        if current_level == 0 and isinstance(color, (list, np.ndarray)):
+            col = color[i]
+        draw_ellipse(center, center, d, d*minor_d_frac, color=col, ax=ax, linewidth=0.5)
+        add_circles(cuts[c_i-1], cuts[c_i+1], 2, col, num_levels, current_level+1, minor_d_frac=minor_d_frac, ax=ax)
+
+
+def plot_cat2cont(ax=None, fontsize='x-large', marker_size=None):
+    #ax = ax or plt.gca()
+
+    major_d = 2
+    minor_d = major_d/4
+
+    ax.set_xlim(0, major_d)
+    ax.set_ylim(0, major_d)
+
+    ellipse_l = (major_d-np.sqrt(2))/2
+    ellipse_r = major_d - ellipse_l
+
+    draw_ellipse(1, 1, major_d*1.1, minor_d, color='black', linewidth=0.5, alpha=0.3, ax=ax)
+
+    colors = ['blue', 'green', 'red']
+    add_circles(ellipse_l, ellipse_r, 3, colors, 3, minor_d_frac=0.75, ax=ax)
+    ax.set_xticks([])
+    ax.set_xticklabels([])
+    ax.set_yticks([])
+    ax.set_yticklabels([])
+
+    p_pos = 0.8
+    p_lw = 4
+    ax.axhline(p_pos, xmax=p_pos/major_d, linewidth=p_lw, c='k')
+    ax.axvline(p_pos, ymax=p_pos/major_d, linewidth=p_lw, c='k')
+    ax.scatter([p_pos], [p_pos], marker='*', color='yellow', edgecolor='black', s=marker_size, alpha=1.0, zorder=10, label="Individual patient")
+    ax.legend(loc='upper left', fontsize=fontsize, frameon=False, borderaxespad=0.1)
+    ax.set_xlabel('Intake features', fontsize=fontsize)
+    ax.set_ylabel('Outcomes', fontsize=fontsize)

activ/viz.py

@@ -92,7 +92,7 @@ def get_nmf_labels():
     # ])
 
     bm_bases_labels = np.array([
-        'Sociomedical factors',
+        'Other sociomedical\nfactors',
         'PEMC',
         'PEMHC',
         'Intracranial injury',

requirements.txt

@@ -5,7 +5,6 @@ h5py
 matplotlib
 seaborn
 umap-learn
-hdbscan
 pandas
 latex
-git+https://github.com/ajtritt/concavity
+git+https://github.com/BouchardLab/concavity