def test_energy(self):
""" Tests function to compute energy statistics"""
self.est.set_params(est='distance', mode='var')
self.assertAlmostEquals(self.est.fit(self.x, biascorr=False),
dc.distance_stats(self.x, self.x)[0])
self.assertAlmostEquals(
self.est.fit(self.x, biascorr=True),
np.sqrt(dc.u_distance_stats_sqr(self.x, self.x)[0]))
self.est.set_params(mode='com')
self.assertAlmostEquals(self.est.fit(self.x, y=self.y, biascorr=False),
dc.distance_covariance(self.x, self.y))
self.est.set_params(mode='mdd')
self.assertAlmostEquals(self.est.fit(self.x, y=self.y, biascorr=False),
0.352427150086)
plt.savefig('SWS_Geolife_ws.png')
plt.show()
dfh,xcl=load_df('Harmonic mean')
plot(dfh,'Harmonic mean',xcl)
for _,__ in dfh.groupby(['Algorithms']):
for k,v in __.groupby('ws'):
print(_,k,v.median().values,v.std().values)
from minepy import MINE
x = __['Harmonic mean'].values
y = __['ws'].values
mine = MINE(alpha=0.6, c=15, est="mic_approx")
mine.compute_score(x, y)
from scipy.spatial import distance
import dcor
print("dcor:",dcor.distance_stats(x, y))
print("distance corrolation:",distance.correlation(x, y))
print("Maximal Information Coefficient :", mine.mic())
print(_,np.corrcoef(__['Harmonic mean'].values,__['ws'].values)[1][0])
r = 0.0
corrxy = 0.0
DIVISOR = 0
for i in range(nsamples):
try:
sample_size = min(1000, len(new_data[el_i]))
new_data_sample = new_data.sample(sample_size)
data_i = pd.to_numeric(new_data_sample[el_i])
data_j = pd.to_numeric(new_data_sample[el_j])
DIVISOR += 1
except ValueError:
continue
ri, pi = pearsonr(data_i, data_j)
covxyi, corrxyi, varxi, varyi = distance_stats(
data_i, data_j)
r += ri
corrxy += corrxyi
r /= DIVISOR
corrxy /= DIVISOR
pearson_corr.append(r)
dist_corr.append(corrxy)
s = "%10s %10s %10s % f % f \n" % (j_, el_i, el_j, r,
corrxy)
fout.write(s)
ax1.plot(dist_corr, pearson_corr, 'o', color=j_colors[j_], label=j_)
fout.close()