def read(Rdiff, gp):
if Rdiff != 'median' and Rdiff != 'min1s' and Rdiff != 'max1s':
print(
'run grd_metalsplit.py to get the split by metallicity done before reading it in for GravImage'
)
exit(1)
import gr_params
gpr = gr_params.grParams(gp)
global Nsample, split, e_split, PM, split_min, split_max
gpr.fil = gpr.dir + "data/tracers.dat"
# number of measured tracer stars
Nsample = bufcount(gpr.fil)
delim = [0, 22, 3, 3, 6, 4, 3, 5, 6, 6, 7, 5, 6, 5, 6, 5, 6]
#ID = np.genfromtxt(gpr.fil,skiprows=29,unpack=True,usecols=(0,1),delimiter=delim)
if gp.case == 5:
RAh, RAm, RAs, DEd, DEm, DEs, VHel, e_VHel, Teff, e_Teff, logg, e_logg, Fe, e_Fe, N = np.loadtxt(
gpr.fil, skiprows=25, unpack=True)
PM = np.ones(len(RAh))
split = logg
e_split = e_logg
sel = (N > 0)
else:
RAh, RAm, RAs, DEd, DEm, DEs, Vmag, VI, VHel, e_VHel, SigFe, e_SigFe, Mg, Mg_err, PM = np.genfromtxt(
gpr.fil,
skiprows=29,
unpack=True,
usecols=tuple(range(2, 17)),
delimiter=delim,
filling_values=-1)
split = Mg
e_split = Mg_err
sel = (Mg > -1) # exclude missing data on Mg
RAh = RAh[sel]
RAm = RAm[sel]
RAs = RAs[sel]
DEd = DEd[sel]
DEm = DEm[sel]
DEs = DEs[sel]
#Vmag = Vmag[sel]
#VI = VI[sel]
VHel = VHel[sel]
e_VHel = e_VHel[sel]
if gp.case < 5:
Mg = Mg[sel]
Mg_err = Mg_err[sel]
elif gp.case == 5:
Teff = Teff[sel]
e_Teff = e_Teff[sel]
logg = logg[sel]
e_logg = e_logg[sel]
Fe = Fe[sel]
e_Fe = e_Fe[sel]
N = N[sel]
split = split[sel]
e_split = e_split[sel]
PM = PM[sel]
split_min = min(split) # -3, 3 if according to WalkerPenarrubia2011
split_max = max(split)
# but: it's not as easy as that
# we have datapoints with errors and probability of membership weighting
# thus, we need to smear the values out using a Gaussian of width = split_err
# and add them up afterwards after scaling with probability PM
x = np.array(np.linspace(split_min, split_max, 100))
splitdf = np.zeros(100)
for i in range(len(split)):
splitdf += PM[i] * gh.gauss(x, split[i], e_split[i])
splitdf /= sum(PM)
sig = abs(RAh[0]) / RAh[0]
RAh = RAh / sig
xs = 15 * (RAh * 3600 + RAm * 60 + RAs) * sig # [arcsec/15]
sig = abs(DEd[0]) / DEd[0]
DEd = DEd / sig
ys = (DEd * 3600 + DEm * 60 + DEs) * sig # [arcsec]
arcsec = 2. * np.pi / (360. * 60. * 60) # [pc]
kpc = 1000 # [pc]
DL = {
1: lambda x: x * (138), #+/- 8 for Fornax
2: lambda x: x * (101), #+/- 5 for Carina
3: lambda x: x * (79), #+/- 4 for Sculptor
4: lambda x: x * (86), #+/- 4 for Sextans
5: lambda x: x * (80) #+/- 10 for Draco
}[gp.case](kpc)
xs *= (arcsec * DL) # [pc]
ys *= (arcsec * DL) # [pc]
# alternative: get center of photometric measurements by deBoer
# for Fornax, we have
if gp.case == 1:
com_x = 96203.736358393697
com_y = -83114.080684733024
xs = xs - com_x
ys = ys - com_y
else:
# determine com_x, com_y from shrinking sphere
import gi_centering as grc
com_x, com_y = grc.com_shrinkcircle_2D(xs, ys)
popass = np.loadtxt(gpr.dir + 'data/popass_' + Rdiff)
sel1 = (popass == 1)
sel2 = (popass == 2)
# radii of all stellar tracers from pop 1 and 2
R1 = np.sqrt((xs[sel1])**2 + (ys[sel1])**2)
R2 = np.sqrt((xs[sel2])**2 + (ys[sel2])**2)
R1.sort()
R2.sort()
R0 = np.hstack([R1, R2])
R0.sort()
for pop in np.arange(2) + 1:
if pop == 1:
Rhalf = R1[len(R1) / 2]
co = 'blue'
else:
Rhalf = R2[len(R2) / 2]
co = 'red'
Rmin = min(R0) # [pc]
Rmax = max(R0) # [pc]
Binmin, Binmax, Rbin = gh.determine_radius(R0, Rmin, Rmax, gp) # [pc]
gp.xipol = Rbin # [pc]
minr = min(Rbin) # [pc]
maxr = max(Rbin) # [pc]
Vol = gh.volume_circular_ring(Binmin, Binmax, gp) # [pc^2]
totmass_tracers = float(len(x))
Rsi = gh.add_errors(R0, gpr.Rerr) # [pc], gpr.Rerr was in
tpb = np.zeros(gp.nipol)
Sig_phot = np.zeros(gp.nipol)
for i in range(gp.nipol):
ind1 = np.argwhere(np.logical_and(Rsi >= Binmin[i],
Rsi < Binmax[i])).flatten() # [1]
tpb[i] = float(len(ind1)) # [1]
Sig_phot[i] = float(
len(ind1)) * totmass_tracers / Vol[i] # [Munit/pc^2]
#loglog(gp.xipol, Sig_phot, co)
#axvline(Rhalf, color=co)
#xlim([min(gp.xipol), max(gp.xipol)])
#xlabel(r'$R$')
#ylabel(r'$\Sigma(R)$')
#pdb.set_trace()
# deproject to get 3D nu profiles
gp.xipol = Rbin
minr = min(Rbin) # [pc]
maxr = max(Rbin) # [pc]
gp.xepol = np.hstack(
[minr / 8., minr / 4., minr / 2., Rbin, 2 * maxr, 4 * maxr,
8 * maxr]) #[pc]
gp.xfine = introduce_points_in_between(gp.xepol, gp)
#pdb.set_trace()
#Sigdatnu, Sigerrnu = gh.complete_nu(Rbin, Sig_phot, Sig_phot/10., gp.xfine)
#dummyx,nudatnu,nuerrnu,Mrnu = gip.Sig_NORM_rho(gp.xfine,Sigdatnu,Sigerrnu,gp)
#nudat = gh.linipollog(gp.xfine, nudatnu, gp.xipol)
#nuerr = gh.linipollog(gp.xfine, nuerrnu, gp.xipol)
#loglog(gp.xipol, nudat, co)
#axvline(Rhalf, color=co)
#xlim([min(gp.xipol), max(gp.xipol)])
#xlabel(r'$R$')
#ylabel(r'$\nu(R)$')
#plum = 100*gh.plummer(gp.xipol, Rhalf, len(R0))
#loglog(gp.xipol, plum, color=co, linestyle='--')
#ylim([min(plum), max(plum)])
#pdb.set_trace()
return
def read(Rdiff, gp):
if Rdiff != 'median' and Rdiff != 'min1s' and Rdiff != 'max1s':
print('run grd_metalsplit.py to get the split by metallicity done before reading it in for GravImage')
exit(1)
import gr_params
gpr = gr_params.grParams(gp)
global Nsample, split, e_split, PM, split_min, split_max
gpr.fil = gpr.dir+"data/tracers.dat"
# number of measured tracer stars
Nsample = bufcount(gpr.fil)
delim = [0,22,3,3,6,4,3,5,6,6,7,5,6,5,6,5,6]
#ID = np.genfromtxt(gpr.fil,skiprows=29,unpack=True,usecols=(0,1),delimiter=delim)
if gp.case==5:
RAh,RAm,RAs,DEd,DEm,DEs,VHel,e_VHel,Teff,e_Teff,logg,e_logg,Fe,e_Fe,N=np.loadtxt(gpr.fil, skiprows=25, unpack=True)
PM = np.ones(len(RAh))
split = logg
e_split = e_logg
sel = (N>0)
else:
RAh,RAm,RAs,DEd,DEm,DEs,Vmag,VI,VHel,e_VHel,SigFe,e_SigFe, Mg,Mg_err,PM = np.genfromtxt(gpr.fil, skiprows=29, unpack=True, usecols=tuple(range(2,17)), delimiter=delim, filling_values=-1)
split = Mg
e_split = Mg_err
sel = (Mg>-1) # exclude missing data on Mg
RAh = RAh[sel]
RAm = RAm[sel]
RAs = RAs[sel]
DEd = DEd[sel]
DEm = DEm[sel]
DEs = DEs[sel]
#Vmag = Vmag[sel]
#VI = VI[sel]
VHel = VHel[sel]
e_VHel = e_VHel[sel]
if gp.case < 5:
Mg = Mg[sel]
Mg_err = Mg_err[sel]
elif gp.case == 5:
Teff = Teff[sel]
e_Teff = e_Teff[sel]
logg = logg[sel]
e_logg = e_logg[sel]
Fe = Fe[sel]
e_Fe = e_Fe[sel]
N = N[sel]
split = split[sel]
e_split = e_split[sel]
PM = PM[sel]
split_min = min(split) # -3, 3 if according to WalkerPenarrubia2011
split_max = max(split)
# but: it's not as easy as that
# we have datapoints with errors and probability of membership weighting
# thus, we need to smear the values out using a Gaussian of width = split_err
# and add them up afterwards after scaling with probability PM
x = np.array(np.linspace(split_min, split_max, 100))
splitdf = np.zeros(100)
for i in range(len(split)):
splitdf += PM[i]*gh.gauss(x, split[i], e_split[i])
splitdf /= sum(PM)
sig = abs(RAh[0])/RAh[0]
RAh = RAh/sig
xs = 15*(RAh*3600+RAm*60+RAs)*sig # [arcsec/15]
sig = abs(DEd[0])/DEd[0]
DEd = DEd/sig
ys = (DEd*3600+DEm*60+DEs)*sig # [arcsec]
arcsec = 2.*np.pi/(360.*60.*60) # [pc]
kpc = 1000 # [pc]
DL = {1: lambda x: x * (138),#+/- 8 for Fornax
2: lambda x: x * (101),#+/- 5 for Carina
3: lambda x: x * (79), #+/- 4 for Sculptor
4: lambda x: x * (86), #+/- 4 for Sextans
5: lambda x: x * (80) #+/- 10 for Draco
}[gp.case](kpc)
xs *= (arcsec*DL) # [pc]
ys *= (arcsec*DL) # [pc]
# alternative: get center of photometric measurements by deBoer
# for Fornax, we have
if gp.case == 1:
com_x = 96203.736358393697
com_y = -83114.080684733024
xs = xs-com_x
ys = ys-com_y
else:
# determine com_x, com_y from shrinking sphere
import gi_centering as grc
com_x, com_y = grc.com_shrinkcircle_2D(xs, ys)
popass = np.loadtxt(gpr.dir+'data/popass_'+Rdiff)
sel1 = (popass==1)
sel2 = (popass==2)
# radii of all stellar tracers from pop 1 and 2
R1 = np.sqrt((xs[sel1])**2 + (ys[sel1])**2)
R2 = np.sqrt((xs[sel2])**2 + (ys[sel2])**2)
R1.sort()
R2.sort()
R0 = np.hstack([R1, R2])
R0.sort()
for pop in np.arange(2)+1:
if pop == 1:
Rhalf = R1[len(R1)/2]
co = 'blue'
else:
Rhalf = R2[len(R2)/2]
co = 'red'
Rmin = min(R0) # [pc]
Rmax = max(R0) # [pc]
Binmin, Binmax, Rbin = gh.determine_radius(R0, Rmin, Rmax, gp) # [pc]
gp.xipol = Rbin # [pc]
minr = min(Rbin)# [pc]
maxr = max(Rbin)# [pc]
Vol = gh.volume_circular_ring(Binmin, Binmax, gp) # [pc^2]
totmass_tracers = float(len(x))
Rsi = gh.add_errors(R0, gpr.Rerr) # [pc], gpr.Rerr was in
tpb = np.zeros(gp.nipol)
Sig_phot = np.zeros(gp.nipol)
for i in range(gp.nipol):
ind1 = np.argwhere(np.logical_and(Rsi >= Binmin[i], Rsi < Binmax[i])).flatten() # [1]
tpb[i] = float(len(ind1)) # [1]
Sig_phot[i] = float(len(ind1))*totmass_tracers/Vol[i] # [Munit/pc^2]
#loglog(gp.xipol, Sig_phot, co)
#axvline(Rhalf, color=co)
#xlim([min(gp.xipol), max(gp.xipol)])
#xlabel(r'$R$')
#ylabel(r'$\Sigma(R)$')
#pdb.set_trace()
# deproject to get 3D nu profiles
gp.xipol = Rbin
minr = min(Rbin) # [pc]
maxr = max(Rbin) # [pc]
gp.xepol =np.hstack([minr/8.,minr/4.,minr/2.,Rbin,2*maxr,4*maxr,8*maxr])#[pc]
gp.xfine = introduce_points_in_between(gp.xepol, gp)
#pdb.set_trace()
#Sigdatnu, Sigerrnu = gh.complete_nu(Rbin, Sig_phot, Sig_phot/10., gp.xfine)
#dummyx,nudatnu,nuerrnu,Mrnu = gip.Sig_NORM_rho(gp.xfine,Sigdatnu,Sigerrnu,gp)
#nudat = gh.linipollog(gp.xfine, nudatnu, gp.xipol)
#nuerr = gh.linipollog(gp.xfine, nuerrnu, gp.xipol)
#loglog(gp.xipol, nudat, co)
#axvline(Rhalf, color=co)
#xlim([min(gp.xipol), max(gp.xipol)])
#xlabel(r'$R$')
#ylabel(r'$\nu(R)$')
#plum = 100*gh.plummer(gp.xipol, Rhalf, len(R0))
#loglog(gp.xipol, plum, color=co, linestyle='--')
#ylim([min(plum), max(plum)])
#pdb.set_trace()
return
def run(gp):
import gr_params
gpr = gr_params.grParams(gp)
global Nsample, split, e_split, PM, split_min, split_max
gpr.fil = gpr.dir + "data/tracers.dat"
# number of measured tracer stars
Nsample = bufcount(gpr.fil)
delim = [0, 22, 3, 3, 6, 4, 3, 5, 6, 6, 7, 5, 6, 5, 6, 5, 6]
#ID = np.genfromtxt(gpr.fil,skiprows=29,unpack=True,usecols=(0,1),delimiter=delim)
if gp.case == 5:
RAh, RAm, RAs, DEd, DEm, DEs, VHel, e_VHel, Teff, e_Teff, logg, e_logg, Fe, e_Fe, N = np.loadtxt(
gpr.fil, skiprows=25, unpack=True)
PM = np.ones(len(RAh))
split = logg
e_split = e_logg
else:
RAh, RAm, RAs, DEd, DEm, DEs, Vmag, VI, VHel, e_VHel, SigFe, e_SigFe, Mg, Mg_err, PM = np.genfromtxt(
gpr.fil,
skiprows=29,
unpack=True,
usecols=tuple(range(2, 17)),
delimiter=delim,
filling_values=-1)
split = Mg
e_split = Mg_err
if gp.case == 5:
sel = (N > 0)
else:
sel = (Mg > -1) # exclude missing data on Mg
RAh = RAh[sel]
RAm = RAm[sel]
RAs = RAs[sel]
DEd = DEd[sel]
DEm = DEm[sel]
DEs = DEs[sel]
#Vmag = Vmag[sel]
#VI = VI[sel]
VHel = VHel[sel]
e_VHel = e_VHel[sel]
if gp.case < 5:
Mg = Mg[sel]
Mg_err = Mg_err[sel]
elif gp.case == 5:
Teff = Teff[sel]
e_Teff = e_Teff[sel]
logg = logg[sel]
e_logg = e_logg[sel]
Fe = Fe[sel]
e_Fe = e_Fe[sel]
N = N[sel]
split = split[sel]
e_split = e_split[sel]
PM = PM[sel]
split_min = min(split) # -3, 3 if according to WalkerPenarrubia2011
split_max = max(split)
# easiest way for visualization: use histogram to show data
#hist(split, np.sqrt(len(split))/2, normed=True)
# but: it's not as easy as that
# we have datapoints with errors and probability of membership weighting
# thus, we need to smear the values out using a Gaussian of width = split_err
# and add them up afterwards after scaling with probability PM
x = np.array(np.linspace(split_min, split_max, 100))
splitdf = np.zeros(100)
for i in range(len(split)):
splitdf += PM[i] * gh.gauss(x, split[i], e_split[i])
splitdf /= sum(PM)
#plot(x, Mgdf, 'g', lw=2)
# only then we want to compare to Gaussians
n_dims = 1 + gp.pops * 2
#Nsample = 10*n_dims
pymultinest.run(
myloglike,
myprior,
n_dims, # nest_ndims
n_dims + 1, # nest_totPar
n_dims, # separate modes on nest_nCdims
# the rho parameters only (gp.nrho in this case)
[gp.pops, gp.nipol, gp.nrho],
True, # nest_IS = INS enabled
True, #nest_mmodal = # separate modes
True, # nest_ceff = use const sampling efficiency
Nsample, # nest_nlive =
0.0, # nest_tol = 0 to keep working infinitely
0.8, # nest_ef =
10000, # nest_updInt = output after this many iterations
1., # null_log_evidence separate modes if
#logevidence > this param.
Nsample, # maxClst =
-1.e30, # nest_Ztol = mode tolerance in the
#case where no special value exists: highly negative
gp.files.outdir, # outputfiles_basename =
-1, # seed =
True, # nest_fb =
False, # nest_resume =
0, # context =
True, # nest_outfile =
-999999, # nest_logZero = points with log L < log_zero will be
1000, # nest_maxIter =
False, # initMPI = use MPI
None) #dump_callback =
import os
os.system(
'cd ' + gp.files.outdir +
'; grep -n6 Maximum stats.dat|tail -5|cut -d " " -f8 > metalmaxL.dat;')
os.system("cd " + gp.files.outdir +
"; sed -i 's/\\([0-9]\\)-\\([0-9]\\)/\\1E-\\2/g' metalmaxL.dat")
os.system("cd " + gp.files.outdir +
"; sed -i 's/\\([0-9]\\)+\\([0-9]\\)/\\1E+\\2/g' metalmaxL.dat")
cubeML = np.loadtxt(gp.files.outdir + 'metalmaxL.dat')
cubeMLphys = cubeML #myprior(cubeML, 1+gp.pops*2, 1+gp.pops*2)
#myloglike(cubeMLphys, 1+gp.pops*2, 1+gp.pops*2)
pML, mu1ML, sig1ML, mu2ML, sig2ML = cubeMLphys
#g1 = pML*gh.gauss(x, mu1ML, sig1ML)
#g2 = (1-pML)*gh.gauss(x, mu2ML, sig2ML)
#gtot = g1+g2
#plot(x, pML*g1, 'white')
#plot(x, (1-pML)*g2, 'white')
#plot(x, gtot, 'r')
#xlabel('Mg')
#ylabel('pdf')
#pdb.set_trace()
sig = abs(RAh[0]) / RAh[0]
RAh = RAh / sig
xs = 15 * (RAh * 3600 + RAm * 60 + RAs) * sig # [arcsec/15]
sig = abs(DEd[0]) / DEd[0]
DEd = DEd / sig
ys = (DEd * 3600 + DEm * 60 + DEs) * sig # [arcsec]
arcsec = 2. * np.pi / (360. * 60. * 60) # [pc]
kpc = 1000 # [pc]
DL = {
1: lambda x: x * (138), #+/- 8 for Fornax
2: lambda x: x * (101), #+/- 5 for Carina
3: lambda x: x * (79), #+/- 4 for Sculptor
4: lambda x: x * (86), #+/- 4 for Sextans
5: lambda x: x * (80) #+/- 10 for Draco
}[gp.case](kpc)
xs *= (arcsec * DL) # [pc]
ys *= (arcsec * DL) # [pc]
# alternative: get center of photometric measurements by deBoer
# for Fornax, we have
if gp.case == 1:
com_x = 96203.736358393697
com_y = -83114.080684733024
xs = xs - com_x
ys = ys - com_y
else:
# determine com_x, com_y from shrinking sphere
import gi_centering as grc
com_x, com_y = grc.com_shrinkcircle_2D(xs, ys)
# instantiate different samplings, store half-light radii (2D)
coll_R1half = []
coll_R2half = []
coll_popass = []
print('drawing 1000 assignments of stars to best fitting Gaussians')
import numpy.random as npr
#import gi_project as gip
for kl in range(1000):
# get a sample assignment:
popass = []
for i in range(sum(sel)):
# random assignment, wrong
#if npr.rand() <= 0.5:
# popass.append(1)
#else:
# popass.append(2)
spl = split[i]
ppop1 = pML * gh.gauss(spl, mu1ML, sig1ML)
ppop2 = (1 - pML) * gh.gauss(spl, mu2ML, sig2ML)
if npr.rand() <= ppop1 / (ppop1 + ppop2):
popass.append(1)
else:
popass.append(2)
popass = np.array(popass)
coll_popass.append(popass)
sel1 = (popass == 1)
sel2 = (popass == 2)
# radii of all stellar tracers from pop 1 and 2
R1 = np.sqrt((xs[sel1])**2 + (ys[sel1])**2)
R2 = np.sqrt((xs[sel2])**2 + (ys[sel2])**2)
R1.sort()
R2.sort()
for pop in np.arange(2) + 1:
if pop == 1:
R0 = R1 # [pc]
Rhalf = R1[len(R1) / 2]
coll_R1half.append(Rhalf)
co = 'blue'
else:
R0 = R2 # [pc]
Rhalf = R2[len(R2) / 2]
coll_R2half.append(Rhalf)
co = 'red'
coll_R1half = np.array(coll_R1half)
coll_R2half = np.array(coll_R2half)
coll_Rdiffhalf = np.abs(coll_R1half - coll_R2half)
# select 3 assignments: one for median, one for median-1sigma, one for median+1sigma
med_Rdiff = np.median(coll_Rdiffhalf)
stdif = np.std(coll_Rdiffhalf)
min1s_Rdiff = med_Rdiff - stdif
max1s_Rdiff = med_Rdiff + stdif
#clf()
#hist(coll_Rdiffhalf, np.sqrt(len(coll_Rdiffhalf))/2)
#xlabel(r'$\Delta R/pc$')
#ylabel('count')
#axvline(med_Rdiff, color='r')
#axvline(min1s_Rdiff, color='g')
#axvline(max1s_Rdiff, color='g')
kmed = np.argmin(abs(coll_Rdiffhalf - med_Rdiff))
kmin1s = np.argmin(abs(coll_Rdiffhalf - min1s_Rdiff))
kmax1s = np.argmin(abs(coll_Rdiffhalf - max1s_Rdiff))
print('saving median, lower 68%, upper 68% stellar assignments')
np.savetxt(gpr.dir + 'data/popass_median', coll_popass[kmed])
np.savetxt(gpr.dir + 'data/popass_min1s', coll_popass[kmin1s])
np.savetxt(gpr.dir + 'data/popass_max1s', coll_popass[kmax1s])
print('finished')
def run(gp):
import gr_params
gpr = gr_params.grParams(gp)
global Nsample, split, e_split, PM, split_min, split_max
gpr.fil = gpr.dir+"data/tracers.dat"
# number of measured tracer stars
Nsample = bufcount(gpr.fil)
delim = [0,22,3,3,6,4,3,5,6,6,7,5,6,5,6,5,6]
#ID = np.genfromtxt(gpr.fil,skiprows=29,unpack=True,usecols=(0,1),delimiter=delim)
if gp.case==5:
RAh,RAm,RAs,DEd,DEm,DEs,VHel,e_VHel,Teff,e_Teff,logg,e_logg,Fe,e_Fe,N=np.loadtxt(gpr.fil, skiprows=25, unpack=True)
PM = np.ones(len(RAh))
split = logg
e_split = e_logg
else:
RAh,RAm,RAs,DEd,DEm,DEs,Vmag,VI,VHel,e_VHel,SigFe,e_SigFe, Mg,Mg_err,PM = np.genfromtxt(gpr.fil, skiprows=29, unpack=True, usecols=tuple(range(2,17)), delimiter=delim, filling_values=-1)
split = Mg
e_split = Mg_err
if gp.case == 5:
sel = (N>0)
else:
sel = (Mg>-1) # exclude missing data on Mg
RAh = RAh[sel]
RAm = RAm[sel]
RAs = RAs[sel]
DEd = DEd[sel]
DEm = DEm[sel]
DEs = DEs[sel]
#Vmag = Vmag[sel]
#VI = VI[sel]
VHel = VHel[sel]
e_VHel = e_VHel[sel]
if gp.case < 5:
Mg = Mg[sel]
Mg_err = Mg_err[sel]
elif gp.case == 5:
Teff = Teff[sel]
e_Teff = e_Teff[sel]
logg = logg[sel]
e_logg = e_logg[sel]
Fe = Fe[sel]
e_Fe = e_Fe[sel]
N = N[sel]
split = split[sel]
e_split = e_split[sel]
PM = PM[sel]
split_min = min(split) # -3, 3 if according to WalkerPenarrubia2011
split_max = max(split)
# easiest way for visualization: use histogram to show data
#hist(split, np.sqrt(len(split))/2, normed=True)
# but: it's not as easy as that
# we have datapoints with errors and probability of membership weighting
# thus, we need to smear the values out using a Gaussian of width = split_err
# and add them up afterwards after scaling with probability PM
x = np.array(np.linspace(split_min, split_max, 100))
splitdf = np.zeros(100)
for i in range(len(split)):
splitdf += PM[i]*gh.gauss(x, split[i], e_split[i])
splitdf /= sum(PM)
#plot(x, Mgdf, 'g', lw=2)
# only then we want to compare to Gaussians
n_dims = 1+gp.pops*2
#Nsample = 10*n_dims
pymultinest.run(myloglike, myprior, n_dims, # nest_ndims
n_dims+1, # nest_totPar
n_dims, # separate modes on nest_nCdims
# the rho parameters only (gp.nrho in this case)
[ gp.pops, gp.nipol, gp.nrho],
True, # nest_IS = INS enabled
True, #nest_mmodal = # separate modes
True, # nest_ceff = use const sampling efficiency
Nsample, # nest_nlive =
0.0, # nest_tol = 0 to keep working infinitely
0.8, # nest_ef =
10000, # nest_updInt = output after this many iterations
1., # null_log_evidence separate modes if
#logevidence > this param.
Nsample, # maxClst =
-1.e30, # nest_Ztol = mode tolerance in the
#case where no special value exists: highly negative
gp.files.outdir, # outputfiles_basename =
-1, # seed =
True, # nest_fb =
False, # nest_resume =
0, # context =
True, # nest_outfile =
-999999, # nest_logZero = points with log L < log_zero will be
1000, # nest_maxIter =
False, # initMPI = use MPI
None) #dump_callback =
import os
os.system('cd '+gp.files.outdir+'; grep -n6 Maximum stats.dat|tail -5|cut -d " " -f8 > metalmaxL.dat;')
os.system("cd "+gp.files.outdir+"; sed -i 's/\\([0-9]\\)-\\([0-9]\\)/\\1E-\\2/g' metalmaxL.dat")
os.system("cd "+gp.files.outdir+"; sed -i 's/\\([0-9]\\)+\\([0-9]\\)/\\1E+\\2/g' metalmaxL.dat")
cubeML = np.loadtxt(gp.files.outdir+'metalmaxL.dat')
cubeMLphys = cubeML #myprior(cubeML, 1+gp.pops*2, 1+gp.pops*2)
#myloglike(cubeMLphys, 1+gp.pops*2, 1+gp.pops*2)
pML, mu1ML, sig1ML, mu2ML, sig2ML = cubeMLphys
#g1 = pML*gh.gauss(x, mu1ML, sig1ML)
#g2 = (1-pML)*gh.gauss(x, mu2ML, sig2ML)
#gtot = g1+g2
#plot(x, pML*g1, 'white')
#plot(x, (1-pML)*g2, 'white')
#plot(x, gtot, 'r')
#xlabel('Mg')
#ylabel('pdf')
#pdb.set_trace()
sig = abs(RAh[0])/RAh[0]
RAh = RAh/sig
xs = 15*(RAh*3600+RAm*60+RAs)*sig # [arcsec/15]
sig = abs(DEd[0])/DEd[0]
DEd = DEd/sig
ys = (DEd*3600+DEm*60+DEs)*sig # [arcsec]
arcsec = 2.*np.pi/(360.*60.*60) # [pc]
kpc = 1000 # [pc]
DL = {1: lambda x: x * (138),#+/- 8 for Fornax
2: lambda x: x * (101),#+/- 5 for Carina
3: lambda x: x * (79), #+/- 4 for Sculptor
4: lambda x: x * (86), #+/- 4 for Sextans
5: lambda x: x * (80) #+/- 10 for Draco
}[gp.case](kpc)
xs *= (arcsec*DL) # [pc]
ys *= (arcsec*DL) # [pc]
# alternative: get center of photometric measurements by deBoer
# for Fornax, we have
if gp.case == 1:
com_x = 96203.736358393697
com_y = -83114.080684733024
xs = xs-com_x
ys = ys-com_y
else:
# determine com_x, com_y from shrinking sphere
import gi_centering as grc
com_x, com_y = grc.com_shrinkcircle_2D(xs, ys)
# instantiate different samplings, store half-light radii (2D)
coll_R1half = []
coll_R2half = []
coll_popass = []
print('drawing 1000 assignments of stars to best fitting Gaussians')
import numpy.random as npr
#import gi_project as gip
for kl in range(1000):
# get a sample assignment:
popass = []
for i in range(sum(sel)):
# random assignment, wrong
#if npr.rand() <= 0.5:
# popass.append(1)
#else:
# popass.append(2)
spl = split[i]
ppop1 = pML*gh.gauss(spl, mu1ML, sig1ML)
ppop2 = (1-pML)*gh.gauss(spl, mu2ML, sig2ML)
if npr.rand() <= ppop1/(ppop1+ppop2):
popass.append(1)
else:
popass.append(2)
popass = np.array(popass)
coll_popass.append(popass)
sel1 = (popass==1)
sel2 = (popass==2)
# radii of all stellar tracers from pop 1 and 2
R1 = np.sqrt((xs[sel1])**2 + (ys[sel1])**2)
R2 = np.sqrt((xs[sel2])**2 + (ys[sel2])**2)
R1.sort()
R2.sort()
for pop in np.arange(2)+1:
if pop == 1:
R0 = R1 # [pc]
Rhalf = R1[len(R1)/2]
coll_R1half.append(Rhalf)
co = 'blue'
else:
R0 = R2 # [pc]
Rhalf = R2[len(R2)/2]
coll_R2half.append(Rhalf)
co = 'red'
coll_R1half = np.array(coll_R1half)
coll_R2half = np.array(coll_R2half)
coll_Rdiffhalf = np.abs(coll_R1half-coll_R2half)
# select 3 assignments: one for median, one for median-1sigma, one for median+1sigma
med_Rdiff = np.median(coll_Rdiffhalf)
stdif = np.std(coll_Rdiffhalf)
min1s_Rdiff = med_Rdiff-stdif
max1s_Rdiff = med_Rdiff+stdif
#clf()
#hist(coll_Rdiffhalf, np.sqrt(len(coll_Rdiffhalf))/2)
#xlabel(r'$\Delta R/pc$')
#ylabel('count')
#axvline(med_Rdiff, color='r')
#axvline(min1s_Rdiff, color='g')
#axvline(max1s_Rdiff, color='g')
kmed = np.argmin(abs(coll_Rdiffhalf-med_Rdiff))
kmin1s = np.argmin(abs(coll_Rdiffhalf-min1s_Rdiff))
kmax1s = np.argmin(abs(coll_Rdiffhalf-max1s_Rdiff))
print('saving median, lower 68%, upper 68% stellar assignments')
np.savetxt(gpr.dir+'data/popass_median', coll_popass[kmed])
np.savetxt(gpr.dir+'data/popass_min1s', coll_popass[kmin1s])
np.savetxt(gpr.dir+'data/popass_max1s', coll_popass[kmax1s])
print('finished')