def test_seds(err=0.1, Av0=1., Z0=0.02):
import grid
#filters = 'hst_wfc3_f225w hst_wfc3_f336w hst_acs_hrc_f475w hst_acs_hrc_f814w hst_wfc3_f110w hst_wfc3_f160w'.upper().split()
#osl = stellib.BaSeL()
oAv = extinction.Cardelli()
g = grid.FileSpectralGrid('libs/SEDs_basel_padovaiso.fits')
lamb = g.lamb#*1e6
#fake DATA
fakein, l, fakesed, fakeerr = getFake(g, Av0, 3.1, err=err)
mask = numpy.zeros(fakesed.shape, dtype=bool)
mask[3] = True
mask[2] = True
Av = numpy.arange(0.,3., 0.1)
r = numpy.empty( (g.seds.shape[0], len(Av)), dtype=float )
with timeit('Likelihood'):
for k in range(len(Av)):
r[:, k] = job(lamb[:], numpy.copy(fakesed), numpy.copy(fakeerr), mask, numpy.copy(g.seds), oAv, Av=Av[k], Rv=3.1)
return g,r, Av, fakein, lamb, fakesed, fakeerr, Av0, Z0
def generate_spectral_grid_from_isochrones(outfile, osl, oiso, Z=0.02):
""" Reinterpolate a given stellar spectral library on to an Isochrone grid
INPUTS:
outfile str fits file to export to
osl stellib.stellib a stellar library
oiso isochrone.Isochrone an isochrone library
Z float metallicity to use
OUTPUTS:
None
only write into outfile
"""
assert(isNestedInstance(osl, stellib.Stellib) )
assert(isNestedInstance(oiso, isochrone.Isochrone) )
specs = numpy.empty( (oiso.data.nrows+1, len(osl.wavelength)), dtype=float )
specs[-1] = osl.wavelength[:]
progress = 0
with timeit('interpolation'):
for k in range(oiso.data.nrows):
if progress < int(100*(k+1)/oiso.data.nrows):
progress = int(100*(k+1)/oiso.data.nrows)
print "progress... %d / 100" % progress
r = numpy.array( osl.interp(oiso.data['logT'][k], oiso.data['logg'][k], Z, oiso.data['logL'][k]) ).T
specs[k,:] = osl.genSpectrum(r)
pyfits.writeto(outfile, specs)
#copy pars
data = {}
for k in oiso.data.keys():
data[k] = oiso.data[k]
pars = mytables.Table(data, name='Reinterpolated stellib grid')
pars.header['stellib'] = osl.source
pars.header['isoch'] = oiso.source
pars.write(outfile, append=True)
def test_specs():
osl = stellib.BaSeL()
oAv = extinction.Cardelli()
#fake DATA
#fakein = 2000 # random between 0 & 4523, no idea what this is :p
idx = osl.grid.where('(Teff >= 3.6) & (Teff <= 3.8) & (logG >= 4.5) & (logG <= 4.7) & (Z == 0.02)')
fakein = idx[0][0]
fakesed = numpy.copy(osl.spectra[fakein,:])
Av0 = 0.1
lamb = osl.wavelength
tau = getFluxAttenuation(oAv, lamb, Av = Av0, Rv = 3.1)
fakesed *= exp(-tau)
#magerr = 0.05
#fakeerr = fakesed * (1. - 10**(-0.4*magerr) )
fakeerr = 0.5 * fakesed
#get Models
# will be replaced by broad-band SEDs but structure will be identical
seds = numpy.copy(osl.spectra)
#idx = osl.grid.where('(Z == 0.02)')
#seds = osl.spectra[idx]
lamb = osl.wavelength
Av = numpy.arange(0,1,0.1)
r = numpy.empty( (seds.shape[0], len(Av)), dtype=float )
with timeit('Likelihood'):
for k in range(len(Av)):
r[:, k] = job(lamb, fakesed, fakeerr, seds, oAv, Av=Av[k], Rv=3.1)
def plot(_r, idx=None):
import pylab as plt
if _r.ndim == 2:
r = _r.sum(1)
else:
r = _r
if idx == None:
idx = numpy.arange(len(r))
n0, bT, bg = numpy.histogram2d(osl.Teff[idx], osl.logg[idx], bins=[25,11])
n , bT, bg = numpy.histogram2d(osl.Teff[idx], osl.logg[idx], bins=[bT,bg], weights=exp(r) )
n0 = n0.astype(float)/n0.sum()
n = n.astype(float)/n.sum()
n1 = numpy.copy(n[:])
ind = n0 > 0.
n[ind] /= n0[ind]
n /= n.sum()
n1 = numpy.ma.masked_where( n0 == 0, n1 )
n = numpy.ma.masked_where( n0 == 0, n )
plt.figure(1, figsize=(10,10))
plt.clf()
ax0 = plt.subplot(221)
ax0.imshow(n1.T, extent=[min(bT), max(bT), min(bg), max(bg)],
vmin=0., vmax=numpy.max([n1.max(),n.max()]),
origin='lower', aspect='auto')
ax0.plot([osl.Teff[fakein]], [osl.logg[fakein]], 'o', mec='#ff0000', mfc='None', mew=2., ms=10.)
ax0.set_xlabel('logT')
ax0.set_ylabel('logg')
ax0.set_xlim(ax0.get_xlim()[::-1])
ax0.set_ylim(ax0.get_ylim()[::-1])
ax0.set_title('Raw: P0(logT,logg) = K')
ax1 =plt.subplot(222, sharex=ax0, sharey=ax0)
ax1.imshow(n.T, extent=[min(bT), max(bT), min(bg), max(bg)],
vmin=0., vmax=numpy.max([n1.max(),n.max()]),
origin='lower', aspect='auto')
ax1.plot([osl.Teff[fakein]], [osl.logg[fakein]], 'o', mec='#ff0000', mfc='None', mew=2., ms=10.)
ax1.set_xlabel('logT')
ax1.set_ylabel('logg')
ax1.set_xlim(ax1.get_xlim()[::-1])
ax1.set_ylim(ax1.get_ylim()[::-1])
ax1.set_title('Corrected: P/P0(logT,logg) = K')
ax2 =plt.subplot(223)
x = 0.5*(bT[1:]+bT[:-1])
#ax2.step( bT[:-1], n1.sum(1), where='pre', lw=2.)
#ax2.step( bT[:-1], n.sum(1), where='pre', lw=2.)
ax2.plot( x, n1.sum(1), lw=2., label='raw')
ax2.plot( x, n.sum(1), lw=2., label='cor')
ylim = ax2.get_ylim()
ax2.vlines([osl.Teff[fakein]], ylim[0],ylim[1])
ax2.set_ylim(ylim)
ax2.set_xlabel('log(Teff)')
ax2.set_ylabel(r'P( data $\mid$ log(Teff) ) P(Teff) / P0(log(Teff))')
ax2.legend(loc=0, frameon=False, borderaxespad=2., prop={'size':14})
ax3 = plt.subplot(224)
x = 0.5*(bg[1:]+bg[:-1])
#ax3.step( bg[:-1], n1.sum(0), where='pre', lw=2.)
#ax3.step( bg[:-1], n.sum(0), where='pre', lw=2.)
ax3.plot( x, n1.sum(0), lw=2.)
ax3.plot( x, n.sum(0), lw=2.)
ylim = ax3.get_ylim()
ax3.vlines([osl.logg[fakein]], ylim[0],ylim[1])
ax3.set_ylim(ylim)
ax3.set_xlabel('log(g)')
ax3.set_ylabel(r'P( data $\mid$ log(g) ) P(log(g)) / P0(log(g))')
figure.theme(ax=ax0)
figure.theme(ax=ax1)
figure.theme(ax=ax2)
figure.theme(ax=ax3)
plot(r)
return r
