def GetCorrelationLoadings(self):
"""
Returns the correlation loadings matrix based on T and inputMatrix.
For variables in inputMatrix with std = 0, the value 0 is written
in the correlation loadings matrix instead of 'Nan' as it should be
(as for example in Matlab)
"""
# Creates empty matrix for correlation loadings
self.correlationLoadings = zeros(
(shape(self.scores)[1], shape(self.originalMatrix)[1]), Float)
# Calculates correlation loadings with formula:
# correlation = cov(x,y)/(std(x)*std(y))
# For each PC in score matrix
for PC in range(shape(self.scores)[1]):
PCscores = self.scores[:, PC]
PCscoresSTD = std(PCscores)
# For each variable/attribute in original matrix (not meancentered)
for var in range(shape(self.originalMatrix)[1]):
origVar = self.originalMatrix[:, var]
origVarSTD = std(origVar)
# If std = 0 for any variable an OverflowError occurs.
# In such a case the value 0 is written in the matrix
try:
self.correlationLoadings[PC, var] = cov(
PCscores, origVar) / (PCscoresSTD * origVarSTD)
except OverflowError:
self.correlationLoadings[PC, var] = 0
return self.correlationLoadings
def GetCorrelationLoadings(self):
"""
Returns the correlation loadings matrix based on T and inputMatrix.
For variables in inputMatrix with std = 0, the value 0 is written
in the correlation loadings matrix instead of 'Nan' as it should be
(as for example in Matlab)
"""
# Creates empty matrix for correlation loadings
self.correlationLoadings = zeros((shape(self.scores)[1], shape(self.originalMatrix)[1]), Float)
# Calculates correlation loadings with formula:
# correlation = cov(x,y)/(std(x)*std(y))
# For each PC in score matrix
for PC in range(shape(self.scores)[1]):
PCscores = self.scores[:, PC]
PCscoresSTD = std(PCscores)
# For each variable/attribute in original matrix (not meancentered)
for var in range(shape(self.originalMatrix)[1]):
origVar = self.originalMatrix[:, var]
origVarSTD = std(origVar)
# If std = 0 for any variable an OverflowError occurs.
# In such a case the value 0 is written in the matrix
try:
self.correlationLoadings[PC, var] = cov(PCscores, origVar) / (PCscoresSTD * origVarSTD)
except OverflowError:
self.correlationLoadings[PC, var] = 0
return self.correlationLoadings
def calcHist(fileName,attr_name,nbins=70):
a=twiss(fileName)
if default==0:
if npy.isreal(a.__dict__[attr_name]):
indx,frq=npy.histogram(a.__dict__[attr_name],bins=nbins)
if default==1:
if a.__dict__[attr_name].typecode()=='d':
hits=Histogram(a.__dict__[attr_name],nbins)
indx=hits[:,0];frq=hits[:,1]
ave=npy.average(a.__dict__[attr_name])
if default==0: stdev=npy.std(a.__dict__[attr_name])
if default==1: stdev=std(a.__dict__[attr_name])
else: print "data is not real"
f=open(path+'/HIST.tfs','w')
f.write('@ NAME %0'+str(len(attr_name))+'s "'+attr_name+'"\n')
f.write('@ '+attr_name+'AVE %le '+str(ave)+'\n')
f.write('@ '+attr_name+'STD %le '+str(ave)+'\n')
f.write("*"+'%5s %5s' % ("INDX", "FREQ")+"\n")
f.write("$"+'%5s %5s' % ("%le", "%le")+"\n")
for j in range(len(indx)):
f.write('%5s %5s' % (indx[j], frq[j])+"\n")
f.close()
