def fit_a_pixel(args):
if len(args) == 6:
tline1, etline1, cont1, tline2, etline2, cont2 = args
elif len(args) == 7:
tline1, etline1, cont1, tline2, etline2, cont2, pb = args
pb.update()
pargrid1 = (cont1*np.exp(-taugrid1) + (1-np.exp(-taugrid1))*texgrid1)
pargrid2 = (cont2*np.exp(-taugrid2) + (1-np.exp(-taugrid2))*texgrid2)
#spec = (1.0-np.exp(-np.array(tau_nu_cumul)))*(Tex-Tbackground)
match, indbest, chi2 = grid_2p_getmatch(tline1+cont1, etline1, pargrid1,
tline2+cont2, etline2, pargrid2)
likelihood = np.exp(-chi2/2.)
def lwt(x, mask=match):
""" Likelihood-weighted mean """
return (x*likelihood)[mask].sum() / likelihood[mask].sum()
def lstd(x, mask=match):
""" Likelihood-weighted stddev """
return (((x-lwt(x,mask))**2*likelihood)[mask].sum() / likelihood[mask].sum() * (mask.sum()/(mask.sum()-1)))**0.5
chi2best = chi2.flat[indbest]
dens_best = densityarr.flat[indbest]
col_best = columnarr.flat[indbest]
temp_best = temparr.flat[indbest]
opr_best = oprarr.flat[indbest]
best = (dens_best, col_best, temp_best, opr_best, chi2best)
if np.count_nonzero(match) == 0:
log.warn(str(args))
warnings.warn("Found no matches. Returning NaNs")
return best, np.nan, np.nan, np.nan, np.nan, np.nan
chi2_all = chi2[match]
dens_all = densityarr[match]
col_all = columnarr[match]
temp_all = temparr[match]
opr_all = oprarr[match]
mins = map(min, (dens_all, col_all, temp_all, opr_all, chi2_all))
maxs = map(max, (dens_all, col_all, temp_all, opr_all, chi2_all))
mean = map(np.mean, (dens_all, col_all, temp_all, opr_all, chi2_all))
std = map(np.std, (dens_all, col_all, temp_all, opr_all, chi2_all))
lweighted = map(lwt, (densityarr, columnarr, temparr, oprarr))
lerror = map(lstd, (densityarr, columnarr, temparr, oprarr))
return best,mins,maxs,mean,std,chi2best,lweighted,lerror
xx,yy,zz = (82-1, 73-1, 107-1)
xx,yy,zz = (111-1, 110-1, 98-1) # lowdens / highstd? EXTREME degeneracy!
xx,yy,zz = (90-1, 110-1, 110-1) # low dens, high *column* std but low *dens* std
cont1 = cont11[yy,xx]
cont2 = cont22[yy,xx]
tline1 = cube1.cube[zz,yy,xx]
tline2 = cube2.cube[zz,yy,xx]
etline1 = cube1.errorcube[zz,yy,xx]
etline2 = cube2.errorcube[zz,yy,xx]
pargrid1 = (cont1*np.exp(-taugrid1) + (1-np.exp(-taugrid1))*texgrid1)
pargrid2 = (cont2*np.exp(-taugrid2) + (1-np.exp(-taugrid2))*texgrid2)
#spec = (1.0-np.exp(-np.array(tau_nu_cumul)))*(Tex-Tbackground)
match, indbest, chi2 = grid_2p_getmatch(tline1+cont1, etline1, pargrid1,
tline2+cont2, etline2, pargrid2)
match11, indbest11, chi2_11 = grid_getmatch(tline1+cont1, etline1, pargrid1)
match22, indbest22, chi2_22 = grid_getmatch(tline2+cont2, etline2, pargrid2)
bw,bz,by,bx = np.unravel_index(indbest, pargrid1.shape)
pl.jet()
pl.clf()
extent = [densityarr.flat[0],densityarr.flat[-1],
columnarr.flat[0],columnarr.flat[-1]]
pl.imshow(pargrid1[bw,bz,:,:], alpha=0.5, extent=extent)
pl.contour(densityarr[bw,bz,:,:],columnarr[bw,bz,:,:],
pargrid2[bw,bz,:,:], levels=np.linspace(tline2-20*etline2+cont2,
tline2+20*etline2+cont2,
10))
