def cfit_model(p,full_model,wavelens):
#S make a chunk of the model shifted by doppler factor z
ipmodel = mb.numconv(full_model.data,gaussian(p))
factor = np.polyval(p[:1:-1],full_model.wavelens)
unbin_model = factor*ipmodel/np.max(ipmodel)
# model = mb.john_rebin(full_model.wavelens,full_model.data,wavelens,p[0])
model = mb.john_rebin(full_model.wavelens,unbin_model,wavelens,p[0])
return model
def cfit_model(p,full_model,wavelens,order):
#S make a chunk of the model shifted by doppler factor z
ipmodel = mb.numconv(full_model.data,gaussian(p))
# model = mb.john_rebin(full_model.wavelens,full_model.data,wavelens,p[0])
#S continuum subtraction
# coeffs = np.polyfit(wavelens,model,5)
#S do we need to add 1? will just come out in the wash. might want to boost
#S the observed though
# cs_model = model - np.polyval(coeffs,wavelens) + 1.
factor = np.polyval(p[:1:-1],full_model.wavelens)
unbinmodel = factor*ipmodel/np.max(ipmodel)
model = mb.john_rebin(full_model.wavelens,full_model.data,wavelens,p[0])
return model
def chunk_func(p,model,star,order,chunk):
z = p['z'].value
ipwidth = p['ipwidth'].value
c0 = p['c0'].value
c1 = p['c1'].value
c2 = p['c2'].value
#S get all inds for the chunk
ckinds = np.arange(star.cklen)+star.skipf+star.cklen*chunk
#S get all wavelens and data for the chunk
ckwaves = star.wavelens[order][ckinds]
ckdata = star.data[order][ckinds]
ckerrs = star.errs[order][ckinds]
#S get the trimmed indices
inds = star.inds['o'+str(order)+'_c'+str(chunk)]
#S get the wavelengths in the model relevant to the order
lowwave = ckwaves.min()-30.
highwave = ckwaves.max()+30.
low = np.where(full_model.wavelens > lowwave)[0]
high = np.where(full_model.wavelens[low] < highwave)[0]
# the relevant model inds for the order
minds = low[high]
# get the model wavelens and ipdata for the order
modwaves = full_model.wavelens[minds]
#S convolve the model over the relevant wavelengths
modck = mb.numconv(full_model.data[minds],gaussian(ipwidth))
#S add some wavelength scaling
factor = np.polyval([c2,c1,c0],modwaves)
unbinmodel = factor*modck/np.max(modck)
#S rebin the convolved model, and
tmodel = mb.john_rebin(modwaves,unbinmodel,ckwaves,z)
return tmodel
def all_err_func(p,model,star,plot_check=False):
z = p['z'].value
amp = p['amp'].value
ipwidth = p['ipwidth'].value
all_errs = np.array([])
all_edge_errs = np.array([])
j=0
skipf = 148
cklen = 300
# print p[0], p[1]
#S convolve the entire model spectrum, want to trim to just relevant wave-
#S lengths, as well as only when the ip changes.
ipdb.set_trace()
full_model.ipdata = mb.numconv(full_model.data,gaussian(amp,ipwidth))
for order in star.fit_orders:
for ck in star.fit_chunks:
#S indices for all the wavelengths in a chunk, for rebinning
#S purposes
ckinds = np.arange(cklen) + skipf + ck*cklen
#S wavelengths we want to rebin for in the chunk
ckwaves = star.wavelens[order][ckinds]
ckdata = star.data[order][ckinds]
ckerrs = star.errs[order][ckinds]
# get the wavelengths in the model relevant to the order
lowwave = ckwaves.min()-10.
highwave = ckwaves.max()+10.
low = np.where(full_model.wavelens > lowwave)[0]
high = np.where(full_model.wavelens[low] < highwave)[0]
# the relevant model inds for the order
minds = low[high]
#S get the model wavelens and ipdata for the order
modwaves = full_model.wavelens[minds]
modck = full_model.ipdata[minds]
#S the index 'zero' point
c0 = p['c0_o'+str(order)+'_c'+str(ck)].value
c1 = p['c1_o'+str(order)+'_c'+str(ck)].value
c2 = p['c2_o'+str(order)+'_c'+str(ck)].value
# j = (order-np.min(star.fit_orders))*5+ck
# temp_params = [p[0],p[1],p[3*j+2],p[3*j+3],p[3*j+4]]
factor = np.polyval([c2,c1,c0],modwaves)
unbinmodel = factor*modck/np.max(modck)
tmodel = mb.john_rebin(modwaves,unbinmodel,ckwaves,z)
inds = blg.inds['o'+str(order)+'_c'+str(ck)]
star.fit_data['o'+str(order)+'_c'+str(ck)] = tmodel[inds]
if plot_check:
#ipdb.set_trace()
plt.plot(ckwaves[inds],ckdata[inds],zorder=2)
plt.plot(ckwaves[inds],tmodel[inds],linewidth=2,zorder=2)
errs = (tmodel[inds] - ckdata[inds])/(ckerrs[inds])#/1.75)
if ck != star.fit_chunks[0]:
edge_errs = (star.fit_data['o'+str(order)+'_c'+str(ck-1)][-1]-\
star.fit_data['o'+str(order)+'_c'+str(ck)][0])
else:
edge_errs = 0.
# all_errs=np.concatenate([all_errs,errs])
all_errs=np.concatenate([all_errs,errs,[edge_errs]])
all_edge_errs=np.concatenate([all_edge_errs,[edge_errs]])
if plot_check:
print('All errors, and the shape')
print all_errs, np.shape(all_errs)
print('All edge errors, and the shape')
print all_edge_errs, np.shape(all_edge_errs)
plt.plot(blg.wavelens[order],blg.data[order],zorder=1)
# plt.errorbar(blg.wavelens[order],blg.data[order],blg.errs[order],
# zorder=1)
# imsave_path = '/Users/samsonjohnson/Desktop/spitzer/blg_ccorr'+\
# '/images0412/'+'orders'+str(min(blg.fit_orders))+'_'+\
# str(max(blg.fit_orders))+'/order'+str(order)+'cklen'+\
# str(cklen)+'.pdf'
# plt.savefig(imsave_path)
plt.show()
if plot_check:
prevpts = 0
prevmax = 0
for order in blg.fit_orders:
for ck in blg.fit_chunks:
x = np.arange(len(blg.inds['o'+str(order)+'_c'+str(ck)]))+\
prevpts
prevpts += len(x)
tdata = blg.fit_data['o'+str(order)+'_c'+str(ck)]
adata = blg.data[order][np.arange(400)+skipf+ck*cklen]\
[blg.inds['o'+str(order)+'_c'+str(ck)]]
errs = blg.errs[order][np.arange(400)+skipf+ck*cklen]\
[blg.inds['o'+str(order)+'_c'+str(ck)]]/1.75
plt.plot(x,prevmax+np.cumsum(((tdata-adata)/errs)**2),\
label=str(order)+' '+str(ck))
# ipdb.set_trace()
prevmax = np.cumsum(((tdata-adata)/errs)**2)[-1] + prevmax
plt.plot(np.arange(prevpts))
plt.axis([0,prevpts+1000,0,prevpts+1000])
plt.legend()
plt.show()
return np.array(all_errs)