##########################################################################
## Set GAP return as NAN
gap_ret = np.where(dates.time == dates[0].time())[0]
ret1[gap_ret, :] = np.NaN
ret2[gap_ret, :] = np.NaN
# Remove the NaNs
NonNan_index = np.logical_not(np.isnan(ret1))
ret1 = ret1[NonNan_index[:, 0], :]
ret2 = ret2[NonNan_index[:, 0], :]
## Final data
dates = dates[NonNan_index[:, 0]]
data = np.concatenate((ret1, ret2), axis=1)
mean = np.mean(data, axis=0)
corr = bMA.get_corrMatrix(data)
cov = bMA.get_covMatrix(data)
############################################################
################# PLOT DATA ###############################
############################################################
import numpy as np
import scipy.stats as stats
from matplotlib import pyplot as plt
cdf = stats.norm.cdf
data = ret1
if (qq_plot):
# Get the histogram and gaussian estimations !
## Scatter plot of the points
if (preprocess_data):
## Remove the NaNs
NonNan_index = np.logical_not(np.isnan(ret1))
ret1 = ret1[NonNan_index[:,0],:]
dates = dates[NonNan_index[:,0]]
#### Apply some shifting if we want crosscorrelations though time
lag_sigmals = 0
if(lag_sigmals != 0 ):
lag_sigmals = ret1[:-lag_sigmals,:]
## Final data
data = ret1
mean = np.mean(data, axis = 0)
corr = bMA.get_corrMatrix(data)
cov = bMA.get_covMatrix(data)
############################################################
################# PLOT DATA ###############################
############################################################
if (perform_EM):
# N,D = Xdata.shape
data = data - np.mean(data, axis = 0)
Xdata = [data]
######## Create the Distribution object ##############################################
#### 1st Gaussian Distribution
Gaussian_d = Cdist.CDistribution(name = "Gaussian");
Gaussian_d.set_distribution("Gaussian")
X_1,X_2 = X2D[:,[0]], X2D[:,[1]]
mu_1, mu_2 = np.mean(Xproj, axis =0)
std_1,std_2 = np.std(Xproj,axis =0)
cov = np.cov(np.concatenate((X_1,X_2),axis = 1).T).T
std_K = 3
## Do stuff now
ax0 = gl.subplot2grid((1,2), (0,0), rowspan=1, colspan=1)
gl.scatter(X_1,X_2, alpha = 0.5, ax = ax0, lw = 4, AxesStyle = "Normal",
labels = ["Original 2D data","X1: %s" %(symbolIDs[AAPL_id]), "X2: %s" %(symbolIDs[GOOGL_id])])
ax0.axis('equal')
################# Draw the error ellipse #################
mean,w,h,theta = bMA.get_gaussian_ellipse_params(X2D, Chi2val = 2.4477)
corr = bMA.get_corrMatrix(X2D)
cov = bMA.get_covMatrix(X2D)
vecs,vals = bMA.get_eigenVectorsAndValues(X2D)
r_ellipse = bMA.get_ellipse_points(mean,w,h,theta)
gl.plot(r_ellipse[:,0], r_ellipse[:,1], ax = ax0, ls = "--",color = "k", lw = 2,
legend = ["Corr: %.2f"%(corr[0,1])])
gl.plot([mean[0], mean[0] + vecs[0,0]*w],
[mean[1], mean[1] + vecs[0,1]*w], ax = ax0, ls = "--",color = "k")
gl.plot([mean[0], mean[0] + vecs[1,0]*h],
[mean[1], mean[1] + vecs[1,1]*h], ax = ax0, ls = "--",color = "k")
gl.plot([mean[0], mean[0] - vecs[0,0]*w],
[mean[1], mean[1] - vecs[0,1]*w], ax = ax0, ls = "--",color = "k")
gl.plot([mean[0], mean[0] - vecs[1,0]*h],
[mean[1], mean[1] - vecs[1,1]*h], ax = ax0, ls = "--",color = "k")