Cluster ANOVA

clusterAnova.py

from rpy2.robjects import pandas2ri
from rpy2.robjects.conversion import localconverter
from rpy2.robjects.packages import importr
import numpy as np
import os 
import pandas as pd
import rpy2
import rpy2.robjects as ro

wd = os.getcwd()

if (os.defpath==".;C:\\bin"):
    os.environ['R_HOME'] = 'C:/Users/User/Documents/R/R-4.1.2'
    os.environ['R_LIBS'] = 'C:/Users/User/Documents/R/R-4.1.2/library'
else:
    os.environ['R_HOME'] = '/mnt/distvol/R/4.0.5/lib64/R/'

pandas2ri.activate()

df = pd.read_csv("https://raw.githubusercontent.com/thistleknot/Python-Stock/master/data/raw/states.csv")
with localconverter(ro.default_converter + pandas2ri.converter):
    r_from_pd_df = ro.conversion.py2rpy(df)

ro.r('''
f <- function(y) {

model <- kmeans(y[,2:ncol(y)], centers = 2)

return(model)
}
''')

r_f = ro.globalenv['f']
d=(r_f(r_from_pd_df))

print(d)
labels = d[0]
print(labels)
clusters = len(d[1])
print(clusters)
centers = d[1]
print(centers)
totss = d[2]
print('totss',totss)
withinss = d[3]
print(withinss)
tot_withinss = d[4]
print('wss',tot_withinss)
betweenss = d[5]
print('bss',betweenss)

#model = KMeans(n_clusters=k, random_state=0, n_init=100).fit(df.iloc[:,1:])

within_ss = []
for n in range(0,clusters):
  #WSS means the sum of distances between the points and the corresponding centroids for each cluster
  data = df[labels==(n+1)].iloc[:,1:]
  within_ss.append(((data - centers[n])**2).sum(1).sum())

WSS = total_within_ss = np.sum(within_ss)
print('wss',total_within_ss)

#sum of ((deviation from variable means) squared)
tot_ss = np.sum(np.sum((df.iloc[:,1:].iloc[:,1:]-df.iloc[:,1:].iloc[:,1:].mean())**2))
print('tot_ss',tot_ss)

cluster_BSS = []
for n in range(0,clusters):
  #sum((variable/column means cluster - variable/column means data)^2)*len(cluster members)
  BSS = np.sum((df[labels==(n+1)].iloc[:,1:].mean()-np.array(np.mean(df.iloc[:,1:])))**2)*len(df[labels==(n+1)].iloc[:,1:])
  #print(BSS)
  cluster_BSS.append(BSS)

BSS = np.sum(cluster_BSS)
print('bss',BSS)

#print(BSS+total_within_ss)

print(tot_ss/totss)
print(WSS/tot_withinss)
#print(tot_ss-betweenss)
#print(betweenss+tot_withinss)
#print(totss-BSS)
print(BSS/betweenss)

Cleaner (sourced from: https://stats.stackexchange.com/questions/81954/ssb-sum-of-squares-between-clusters)

import numpy as np
from scipy.cluster.vq import vq

X = np.array(df.iloc[:,1:])
codebook = np.array(centers)
partition, euc_distance_to_centroids = vq(X, codebook)
WSS = np.sum(euc_distance_to_centroids**2)

TSS = np.sum((X-X.mean(0))**2)

BSS = TSS - WSS

print(TSS, WSS, BSS)