Added randomized partitioning

HOWTO.md

@@ -55,8 +55,9 @@ Currently, the following standard processes are supported:
 - Visualize the most distinctive words as a horizontal bar chart. (pyzeta.plot_scores)
 - Visualize the feature distribution as a scatterplot (pyzeta.plot_types)
 
-The following experimental functions are present but not really supported:
+The following experimental functions are present (but not really supported):
 
+- If you use ["random", "0", "1"] as the value for the `contrast` parameter, the partitions will be built randomly, splitting the collection in equal-sized parts. This is interesting if you want to see how strong your zeta scores really are relative to a random partitioning. (Expanding on this principle could be the basis for some type of significance test for zeta scores.)
 - Visualize the relation between three partitions based on type proportions in two partitions (pyzeta.threeway)
 - PCA for three partitions using distinctive features.
 

pyzeta.py

@@ -27,6 +27,7 @@
 import itertools
 import shutil
 from sklearn.decomposition import PCA
+import random
 
 
 # =================================
@@ -104,11 +105,17 @@ def make_filelist(metadatafile, contrast):
     with open(metadatafile, "r") as infile:
         metadata = pd.DataFrame.from_csv(infile, sep=";")
         # print(metadata.head())
-        onemetadata = metadata[metadata[contrast[0]].isin([contrast[1]])]
-        twometadata = metadata[metadata[contrast[0]].isin([contrast[2]])]
-        onelist = list(onemetadata.loc[:, "idno"])
-        twolist = list(twometadata.loc[:, "idno"])
-        # print(onelist, twolist)
+        if contrast[0] != "random": 
+            onemetadata = metadata[metadata[contrast[0]].isin([contrast[1]])]
+            twometadata = metadata[metadata[contrast[0]].isin([contrast[2]])]
+            onelist = list(onemetadata.loc[:, "idno"])
+            twolist = list(twometadata.loc[:, "idno"])
+        elif contrast[0] == "random":
+            idnolist = list(metadata.loc[:, "idno"])
+            newidnolist = random.sample(idnolist, len(idnolist))
+            onelist = newidnolist[:int(len(idnolist)/2)]
+            twolist = newidnolist[int(len(idnolist)/2):]
+        print(onelist, twolist)
         print("----number of texts: " + str(len(onelist)) + " and " + str(len(twolist)))
         return onelist, twolist