def downhill(prob_func,namefree,parmfree,namefix,parmfix,data,eventinfo,parmref,isspitz,use_color=False,I_minus_L=[]):
start = time.clock()
tbegin,talert,t0par,t0base,mjd = parmref
alpha,delta,oblats,oblongs,obgammas = eventinfo
## compute (qn,qe) once and for all ##
qnqe = []
for iob in range(len(data)):
date = data[iob][0]
qnqe_iob = genlc.getqnqe(date,alpha,delta,t0par,oblats[iob],oblongs[iob],isspitz[iob])
qnqe.append(qnqe_iob)
### Get chi2 based on initial parameters ###
# chi2_obs,dof,fsfb = prob_func(parmfree,parmfix,namefree,namefix,data,eventinfo,t0par,isspitz,True,use_color,I_minus_L,qnqe,get_chi2_obs=True)
# print chi2_obs,dof
# parmbest = fmin(prob_func,parmfree,args=(parmfix,namefree,namefix,data,eventinfo,t0par,isspitz,True,use_color,I_minus_L))
parmbest,chi2min,iter,funcalls,warnflag,allevcs = fmin(prob_func,parmfree,args=(parmfix,namefree,namefix,data,eventinfo,t0par,isspitz,True,use_color,I_minus_L,qnqe),full_output=True,retall=True,maxiter=3000,maxfun=10000)
print 'best parameters: '
print namefree
print ['%.3f'%iparm for iparm in parmbest]
def mcmc(prob_func,eventname,namefree,parmfree,namefix,parmfix,data,eventinfo,parmref,isspitz,use_color=False,I_minus_L=[]):
start = time.clock()
tbegin,talert,t0par,t0base,mjd = parmref
alpha,delta,oblats,oblongs,obgammas = eventinfo
## compute (qn,qe) once and for all ##
qnqe = []
for iob in range(len(data)):
date = data[iob][0]
qnqe_iob = genlc.getqnqe(date,alpha,delta,t0par,oblats[iob],oblongs[iob],isspitz[iob])
qnqe.append(qnqe_iob)
### Get chi2 based on initial parameters ###
# chi2_obs,dof,fsfb = prob_func(parmfree,parmfix,namefree,namefix,data,eventinfo,t0par,isspitz,True,use_color,I_minus_L,qnqe,get_chi2_obs=True)
# print chi2_obs,dof
ndim = len(parmfree)
nwalkers = 30
pos = [parmfree+1e-3*np.random.randn(ndim) for i in range(nwalkers)]
sampler = emcee.EnsembleSampler(nwalkers,ndim,prob_func,args=(parmfix,namefree,namefix,data,eventinfo,t0par,isspitz,False,use_color,I_minus_L,qnqe),threads=1)
pos,lnprob,rstate = sampler.run_mcmc(pos,100) ## burn in
sampler.reset()
sampler.run_mcmc(pos,200) ## MCMC sampling
## save the MCMC chain and other useful information ##
chain = sampler.chain.reshape((-1,ndim),order='F')
chi2s = -2*sampler.lnprobability.reshape(-1,order='F')
headerstr = '%s '%namefree[0]
for i in range(1,len(namefree)):
headerstr += '%s '%namefree[i]
np.savetxt('output/mcmc/chain-%s.dat'%eventname,chain,fmt='%f',header=headerstr)
np.savetxt('output/mcmc/acceptance-%s'%eventname,sampler.acceptance_fraction,fmt='%f')
np.savetxt('output/mcmc/chisq-%s.dat'%eventname,chi2s)
## find the best-fit and error bars ##
parmbest = chain[chi2s==min(chi2s)][0]
parmmean = np.median(chain,axis=0)
parmerrs = np.std(chain,axis=0)
print 'best parameters: '
print namefree
print ['%.4f'%iparm for iparm in parmbest]
