def fitMassMC(tzero,period,oc,epoch,weight):
z = numpy.polyfit(epoch, oc, 2) # the traditional chi-square fit
print "The chi-square result: ", z
#priors
sig = pymc.Uniform("sig", 0.0, 100.0, value=1.)
a = pymc.Uniform("a", -10.0, 10.0, value= 0.0)
b = pymc.Uniform("b", -10.0, 10.0, value= 0.0)
c = pymc.Uniform("c", -10.0, 10.0, value= 0.0)
#model
@pymc.deterministic(plot=False)
def mod_quadratic(x=epoch, a=a, b=b, c=c):
return a*x**2 + b*x + c
#likelihood
quad_y = pymc.Normal("y", mu=mod_quadratic, tau=1.0/sig**2, value=oc, observed=True)
quad_model = pymc.Model([quad_y, a, b, c])
quadratic_mcmc = pymc.MCMC(quad_model)
quadratic_mcmc.sample(1000000, 200)
#pymc.Matplot.plot(quadratic_mcmc)
a = quadratic_mcmc.trace('a')[:]
b = quadratic_mcmc.trace('b')[:]
c = quadratic_mcmc.trace('c')[:]
print('Last value of a: {}, and the median value is {}'.format(a[-1], numpy.median(a)))
print('Last value of b: {}, and the median value is {}'.format(b[-1], numpy.median(b)))
print('Last value of c: {}, and the median value is {}'.format(c[-1], numpy.median(c)))
print quadratic_mcmc.stats()
xmin = min(epoch)
xmax = max(epoch)
xlen = len(epoch)
fitx = numpy.linspace(xmin,xmax,xlen)
fity = model(a[-1],b[-1],c[-1],fitx)
fitPlot(epoch, oc, minType, method,fitx, fity)
def fitLiteLM(self, userA, ecc, w, tzero, per, npoc, npepoch, weight, limitations):
l = Lite()
l.parameterSummary()
l["a"] = userA
l["ecc"] = ecc
l["w"] = w
l["t0"] = tzero
l["per"] = per
freedomPars = []
if limitations[0] == False:
freedomPars.append("t0")
if limitations[1] == False:
freedomPars.append("per")
if limitations[2] == False:
freedomPars.append("a")
if limitations[3] == False:
freedomPars.append("ecc")
if limitations[4] == False:
freedomPars.append("w")
print freedomPars
l.thaw(freedomPars) # free parameters
print l.thaw
l.fit(npepoch, npoc, yerr=weight)
l.parameterSummary()
model = l.evaluate(npepoch)
sepoch, fity = zip(*sorted(zip(npepoch, model)))
fitPlot(npepoch, npoc, minType, method, sepoch, fity)
def fitLiteLM(self,userA,ecc,w,tzero,per,npoc,npepoch,weight,limitations):
l = Lite()
#l.parameterSummary()
l["a"] = userA
l["ecc"] = ecc
l["w"]=w
l["t0"]=tzero
l["per"]=per
freedomPars = []
if limitations[0] == False:
freedomPars.append("t0")
if limitations[1] == False:
freedomPars.append("per")
if limitations[2] == False:
freedomPars.append("a")
if limitations[3] == False:
freedomPars.append("ecc")
if limitations[4] == False:
freedomPars.append("w")
################## FIT MCMC #####################################
# l.thaw(freedomPars) #free parameters
# l.fit(npepoch,npoc,yerr=weight)
#
# X0 = {"a":l["a"],"ecc":l["ecc"],"w":l["w"], "t0":l["t0"], "per":l["per"]}
# Lims = {"a":[-10.,10.],"ecc":[-1.,1], "w":[-360.,360.], "t0":[0.,50000.], "per":[0.,100000.]}
# steps = {"a":0.01, "ecc":0.01,"w":0.1, "t0":200, "per":500}
#
# l.fitMCMC(npepoch, npoc, X0, Lims, steps, yerr=weight, \
# iter=2500, burn=0, thin=1, \
# dbfile="mcmcExample.tmp")
#------------------------------------------------------------------
# specific limits for MCMC paramters
# ppa = {}
# ppa["a"] = pymc.Uniform("a", value=l["a"], lower=-1., \
# upper=1.0, doc="Amplitude")
# ppa["ecc"] = pymc.Uniform("ecc", value=l["ecc"], lower=0., \
# upper=1.0, doc="Eccentricity")
# Use this for ppa
# l.fitMCMC(npepoch, npoc, X0, Lims, steps, yerr=weight, \
# pymcPars=ppa,iter=2500, burn=0, thin=1, \
# dbfile="mcmcExample.tmp")
##################################################################
######## AUTO FIT is WORKING #####################################
# ranges = {"a":[-10.,10.],"ecc":[0.,1], "w":[0.,360.], "t0":[0.,50000.], "per":[0.,50000.]}
# l.autoFitMCMC(npepoch, npoc, ranges, yerr=weight, iter=1000, picky=False)
##################################################################
######## emcee ###################################################
l.thaw(freedomPars) #free parameters
sampleArgs = {"iters":1000, "burn":200}
priors = {"a":fuf.FuFPrior("limuniform", lower=-10.0, upper=10.), \
"ecc":fuf.FuFPrior("limuniform", lower=-1.0, upper=1.), \
"w":fuf.FuFPrior("limuniform", lower=-360.0, upper=360.), \
"t0":fuf.FuFPrior("limuniform", lower=0.0, upper=50000.), \
"per":fuf.FuFPrior("limuniform", lower=0.0, upper=50000.)}
# Note that the filename should end in .emcee. Substitute this filename
# in the following examples.
l.fitEMCEE(npepoch, npoc, yerr=weight, sampleArgs=sampleArgs, \
dbfile="mcmcTA.emcee", priors=priors)
##################################################################
l.parameterSummary()
model = l.evaluate(npepoch)
sepoch, fity = zip(*sorted(zip(npepoch, model)))
fitPlot(npepoch, npoc, minType, method, sepoch,fity)
#print l1, l2, l3, l4, l5, l6, l7, l8, l9, l10,l11, l12
#return l1, l2, l3, l4, l5, l6, l7, l8, l9, l10,l11, l12, fitx, fity
try:
fitobj.fit(params0=paramsinitial)
except Exception, mes:
print "Something went wrong with fit: ", mes
raise SystemExit
# results
l1 = "Fit status: ", fitobj.message
l2 = "Best-fit parameters: ", fitobj.params
l3 = "Covariance errors: ", fitobj.xerror
l4 = "Standard errors ", fitobj.stderr
l5 = "Chi^2 min: ", fitobj.chi2_min
l6 = "Reduced Chi^2: ", fitobj.rchi2_min
l7 = "Iterations: ", fitobj.niter
l8 = "Number of function calls: ", fitobj.nfev
l9 = "Number of free pars.: ", fitobj.nfree
l10 = "Degrees of freedom: ", fitobj.dof
l11 = "Number of pegged pars.: ", fitobj.npegged
l12 = "Covariance matrix:\n", fitobj.covar
xmin = min(epoch)
xmax = max(epoch)
xlen = len(epoch)
fitx = numpy.linspace(xmin, xmax, xlen)
fity = model(fitobj.params, fitx)
fitPlot(epoch, oc, minType, method, fitx, fity)
return l1, l2, l3, l4, l5, l6, l7, l8, l9, l10, l11, l12, fitx, fity
def fitMassMC(self,tzero,per,npoc,npepoch,weight):
l = massTransfer()
#l.parameterSummary()
# l["a"] = 0
# l["b"] = 0
# l["c"]= 0
################## FIT MCMC #####################################
# l.thaw(["a","b","c"]) #free parameters
# l.fit(npepoch,npoc,yerr=weight)
#
# X0 = {"a":l["a"],"b":l["b"],"c":l["c"]}
# Lims = {"a":[-1e-20,1e20],"b":[-1e-20,1e20], "c":[-1e-20,1e20]}
# steps = {"a":0.01, "b":0.01,"c":0.1}
#
# l.fitMCMC(npepoch, npoc, X0, Lims, steps, yerr=weight, \
# iter=1000, burn=0, thin=1, \
# dbfile="mcmcExample.tmp")
#------------------------------------------------------------------
# specific limits for MCMC paramters
# ppa = {}
# ppa["a"] = pymc.Uniform("a", value=l["a"], lower=-1., \
# upper=1.0, doc="Amplitude")
# ppa["ecc"] = pymc.Uniform("ecc", value=l["ecc"], lower=0., \
# upper=1.0, doc="Eccentricity")
# Use this for ppa
# l.fitMCMC(npepoch, npoc, X0, Lims, steps, yerr=weight, \
# pymcPars=ppa,iter=2500, burn=0, thin=1, \
# dbfile="mcmcExample.tmp")
##################################################################
######## AUTO FIT is WORKING #####################################
# ranges = {"a":[-1e-20,1e20],"b":[-1e-20,1e20], "c":[-1e-20,1e20]}
# l.autoFitMCMC(npepoch, npoc, ranges, yerr=weight, iter=100000, picky=False)
##################################################################
######## emcee ###################################################
l.thaw(["a","b","c"]) #free parameters
sampleArgs = {"iters":100000, "burn":200}
priors = {"a":fuf.FuFPrior("limuniform", lower=-1e-20, upper=1e-20), \
"b":fuf.FuFPrior("limuniform", lower=-1e-20, upper=1e-20), \
"c":fuf.FuFPrior("limuniform", lower=-1e-20, upper=1e-20)}
# Note that the filename should end in .emcee. Substitute this filename
# in the following examples.
l.fitEMCEE(npepoch, npoc, yerr=weight, sampleArgs=sampleArgs, \
dbfile="mcmcTA.emcee", priors=priors)
##################################################################
l.parameterSummary()
model = l.evaluate(npepoch)
sepoch, fity = zip(*sorted(zip(npepoch, model)))
fitPlot(npepoch, npoc, minType, method, sepoch,fity)
