I'm plotting the relative magnitude of the error bars of different sources of uncertainty in the emu inference.
In [1]:
import matplotlib
#matplotlib.use('Agg')
from matplotlib import pyplot as plt
%matplotlib inline
import seaborn as sns
sns.set()
In [2]:
import numpy as np
In [3]:
%%bash
ls /home/users/swmclau2/Git/pearce/bin/mcmc/*scov*.npy -lt
In [4]:
boxno, realization = 3, 1
In [5]:
shot_cov = np.loadtxt('/home/users/swmclau2/Git/pearce/bin/mcmc/xi_gg_shot_cov_true_%d%d.npy'%(boxno, realization))
jk_cov = np.loadtxt('/home/users/swmclau2/Git/pearce/bin/mcmc/xi_gg_cov_true_jk_%d%d.npy'%(boxno, realization))
sample_cov = np.loadtxt('/home/users/swmclau2/Git/pearce/bin/mcmc/xigg_scov.npy')
data_cov = np.loadtxt('xigg_ycov.npy')
In [6]:
#TODO add emu1 cov,
In [11]:
last_bar = np.zeros((shot_cov.shape[0],))
i = 0
for covmat, label in zip([shot_cov, jk_cov], ['Shot', 'JK']):
#if i == 2:
# break
#i+=1
errs = np.sqrt(np.diag(covmat))
print label, errs
if label == 'Training':
plt.bar(np.array(range(shot_cov.shape[0])), errs, bottom = last_bar, label = label, color = 'r')
else:
plt.bar(np.array(range(shot_cov.shape[0])), errs, bottom = last_bar, label = label)
last_bar = errs
plt.legend(loc = 'best')
plt.yscale('log')
plt.show()
In [ ]:
In [ ]: