def uncorrectUncertainty(self,correction=None):
"""
Undoes correction on measurement uncertainty.
correction: Information on how to perform the correction.
May be a path to a pickled file, a float, or
list of values.
"""
if isinstance(correction,(str)):
correction = acs.pklread(correction)
if isinstance(correction,(float,int)):
self.spectra_errs /= np.sqrt(correction)
elif isinstance(correction,(list)):
correction = np.array(correction)
if isinstance(correction,(np.ndarray)):
if correction.shape != self.spectra_errs.shape:
correction = np.tile(correction,(self.spectra_errs.shape[0],1))
self.spectra_errs = np.sqrt(self.spectra_errs**2/correction)
def comp_R2(ms,direc=None,savename=None,
labels = ['raw sample,','corrected \nsample,','M.A.D.,']):
"""
Plot comparison of R^2 arrays for different models.
ms: Either list of model objects or list of strings where
models are stored.
direc: Directory to find model files if ms a list of strings
labels: Label corresponding to both models
"""
# Start figure
plt.figure(figsize = (10,8))
# Find equally spaced colours from the plasma colourmap
colors = plt.get_cmap('plasma')(np.linspace(0, 0.8, len(ms)))
# Choose linestypes and markertypes
types = ['o','s','^']
start = 0
vecs = np.array([],dtype=int)
# For each model
t=0
for i in range(len(ms)):
if t >= len(types):
t = 0
# Construct latex label
r2noise_label = '\n$\mathbf{R^2_{noise}}$ = '
# Find model - if ms[i] a string, read from file, otherwise use entry
if isinstance(ms[i],str):
m = acs.pklread(direc+'/'+ms[i])
else:
m = ms[i]
# Set consistent plot boundaries
plt.ylim(0,1)
plt.xlim(-1,len(m.R2Array))
# Find where R^2 values cross the R^2_noise line
crossvec = np.where(m.R2Array > m.R2noise)
if crossvec[0] != []:
crossvec = crossvec[0][0]-1
if crossvec < 0:
crossvec=0
plt.axvline(crossvec,0,m.R2Array[crossvec],color=colors[i],
linestyle='-',lw=2)
vecs = np.append(vecs,crossvec)
nummarker = np.min([len(m.R2Array),10])
if nummarker < len(m.R2Array):
vec_points = np.arange(len(m.R2Array))[start::len(m.R2Array)/10]
R2_points = m.R2Array[start::len(m.R2Array)/10]
start += len(m.R2Array)/(10*len(ms))
else:
vec_points = np.arange(len(m.R2Array))
R2_points = m.R2Array
# Plot R^2 as a function of number of eigenvectors
plt.plot(m.R2Array,color=colors[i],ls='-',lw=3)
plt.plot(vec_points,R2_points,ls='',ms=14,marker=types[t],markeredgecolor='k',
markeredgewidth=2,markerfacecolor=colors[i],
label = r'{0} {1} {2:.2f}'.format(labels[i],r2noise_label,
m.R2noise))
plt.axhline(m.R2noise,color=colors[i],ls='--',lw=3)
t+=1
# Label axes and add the legend
step=10**np.floor(np.log10(len(m.R2Array)))
#ticklist = np.concatenate((np.arange(0,len(m.R2Array),step,dtype=int),vecs))
#plt.xticks(ticklist,ticklist.astype(str))
for vec in vecs:
plt.text(vec+0.01*len(m.R2Array),0.02,'{0}'.format(vec))
plt.ylabel(r'$\mathbf{R}^2$',fontsize=font['size']+2)
plt.xlabel('n',fontsize=font['size']+2)
legend = plt.legend(loc='best',fontsize=font['size']-2)
legend.get_frame().set_linewidth(0.0)
if not savename:
savename = '{0}/R2_comp.png'.format(direc)
plt.savefig(savename)
