def _mean_vector(self, data):
"returns mean of vector with same fromid and toid"
# add vectors to list
lvec = [] # helper list of vectors
for c in data:
for v in c:
lvec.append(v)
datamean = ObsClusterList()
dmcluster = ObsClusterVector(covmat=CovMat(dim=0, band=2))
datamean.append_cluster(dmcluster)
# list of vector means and single vectors (without mean)
cll_mean = ObsClusterList()
cll_obs = ObsClusterList()
cll_res = ObsClusterList()
if len(lvec) <= 1:
return # nothink to compare
while len(lvec) > 1:
# find vectors with same fromid and toid
v1 = lvec.pop()
fromid = v1.fromid
toid = v1.toid
lvecm = [v1] # list of vectors for mean
lsign = [1.0] # list of signs: for revers vectors -1.0
for v2 in lvec:
if v2.fromid == fromid and v2.toid == toid:
lvecm.append(v2)
lvec.remove(v2)
lsign.append(1.0)
continue
if self.revers and\
v2.fromid == toid and v2.toid == fromid:
lvecm.append(v2)
lvec.remove(v2)
lsign.append(-1.0)
# compute mean
if float(len(lvecm)) == 1:
# nothink to compute
vec = lvecm[0]
covmat = vec.covmat
dmcluster.append_obs(vec)
dmcluster.covmat.extend_dim(3)
vec.covmat = covmat
continue
from numpy import mat, mean
ll = [mat([i.dx, i.dy, i.dz])*sign for i,sign in zip(lvecm,lsign)]
# row vectors of observ.
#lm = mean(ll, axis=0)
#print "Mean 0:", lm
#print "covmat0:", lvecm[0].covmat.mat
# compute covariance matrix
# covariances between vectors not handled
sil = [i.covmat.mat.I for i in lvecm]
# covariance matrice inverses inv(Sigma)
sill = [si*l.T for si,l in zip(sil,ll)]
# product inv(Sigma)*l
cmm = sum(sil).I # covariance matrix of mean
mn = cmm * sum(sill) # mean value
#print "Mean 1:", mn
#print "covmat1:", cmm
# append clusters
cl_mean = ObsClusterVector(covmat=CovMat(dim=3, band=2))
cl_obs = ObsClusterVector(covmat=CovMat(dim=0, band=2))
cl_res = ObsClusterVector(covmat=None)
cll_mean.append_cluster(cl_mean)
cll_obs.append_cluster(cl_obs)
cll_res.append_cluster(cl_res)
# append vector mean
vec1 = ObsVector(fromid=lvecm[0].fromid, toid=lvecm[0].toid,
dx=float(mn[0]), dy=float(mn[1]), dz=float(mn[2]))
vec2 = ObsVector(fromid=lvecm[0].fromid, toid=lvecm[0].toid,
dx=float(mn[0]), dy=float(mn[1]), dz=float(mn[2]))
# vec1 == vec2
cl_mean.append_obs(vec1)
dmcluster.append_obs(vec2)
# set covmat of mean
covmat = CovMat(dim=3, band=2)
covmat.mat = cmm
vec1.covmat = covmat
dmcluster.covmat.extend_dim(3)
vec2.covmat = covmat
# append observations with covmat
for v in lvecm:
vcovmat = v.covmat
cl_obs.append_obs(v)
cl_obs.covmat.extend_dim(3)
v.covmat = vcovmat
# append residuals
for l,v,s in zip(ll,lvecm,lsign):
res = (mn - l.T) * s
vec = ObsVector(fromid=v.fromid, toid=v.toid,
dx=float(res[0]),
dy=float(res[1]),
dz=float(res[2]))
cl_res.append_obs(vec)
return datamean, cll_mean, cll_obs, cll_res
cl.textTable = obs_vector_table()
print cl
cl.covmat = CovMat(6,5)
#cl.covmat.stdev = (0.01, 0.02, 0.03, 0.04, 0.05, 0.06)
#cl.covmat.var = (0.01, 0.02, 0.03, 0.04, 0.05)
cl.covmat.data = [[1,2,3,4,5,6], [2,3,4,5,6], [3,4,5,6], [4,5,6], [5,6], [6]]
print cl.make_gama_xml()
cl.textTable = obs_vector_stdev_table()
print cl
# iterator
print "\n".join(["from=%s to=%s" % (obs.fromid, obs.toid) for obs in cl])
#covmat of vectors
print v1.covmat.data
print v2.covmat.data
cm = CovMat(3,2)
cm.data = [[0.1, 0.2, 0.3],[0.2, 0.3],[0.3]]
v1.covmat = cm
print cl.covmat.data
# transformation
#from gizela.tran.Tran3D import Tran3D
#tr = Tran3D()
#from math import pi
#tr.rotation_xyz(pi/2, pi/2, pi/2)
#cl.tran_3d(tr)
#print cl