Ok, so this should make a distance-correlation matrix of everything

Unfortunately, it would take approximately a week on an r5.large, so
I'm going to work with a large sample instead

py/clustering.py

@@ -20,7 +20,9 @@ def tqcorr(df):
     print(cdf.info())
     comb = combinations(cols, 2)
     for c1, c2 in tqdm(comb):
-        cdf.loc[c1, c2] = 1 - df[c1].corr(df[c2])
+        dv = 1 - df[c1].corr(df[c2])
+        cdf.loc[c1, c2] = dv
+        cdf.loc[c2, c1] = dv
     print(cdf.info())
     return cdf
 
@@ -28,9 +30,9 @@ def tqcorr(df):
 tqdm.pandas()
 
 Sourcedata =   '../data/2017-all-wide.pkl'
-lableddata =   '../data/9-clusters.pkl'
-aggdata =      '../data/9-clusters.agg.pkl'
-clustertable = '../data/9-clusters-sample-table.pkl'
+lableddata =   '../data/9-clusters-all.pkl'
+aggdata =      '../data/9-clusters-all-agg.pkl'
+clustertable = '../data/9-clusters-all-table.pkl'
 
 numclusts = 9
 df = p.read_pickle(Sourcedata)