I'm working on a project were I'm performing a model fitting and my problem is that I have 2 of 4 variables that are discrete (this means that are already generated models that I'm just able to read), so I need that when I use the emcee.EnsembleSampler() takes the values of this two variables from 2 list (e.g. m=[1,2,3,4] ,b=[16,17,18,20] ) while the other variables could remains as a continuous variables.
(to get something more concise let say that I'm adjusting a line to a pair of dots where the lines are pre-made in function of the slope m and the position coefficient b, and that a have 2 other variables that let me shift the model in both directions )
# Packages
import numpy as np
impor emcee
# Data
x = x
y = y
yerr = yerr
# Functions
def lnlike(self,theta, x, y, yerr): # Log Likelihood
m, b, xs, ys = theta
model = np.genfromtxt('Generic_name_'+str(m)+str(b)) # a two column file
model[:,0]+=xs
model[:,1]+=ys
inv_sigma2 = 1.0/(yerr**2 )
return -0.5*(np.sum((y-model)**2))
def lnprior(self,theta): # Priors
m, b, xs, ys = theta
if m_min< m <m_max and b_min< b <b_max and xs_min< xs <xs_max and ys_min< ys <ys_max :
return 0.0
return -np.inf
def lnprob(self,theta, x, y, yerr): # Log Probability
lp = self.lnprior(theta)
if not np.isfinite(lp):
return -np.inf
return lp + self.lnlike(theta, x, y, yerr) # This return the posterior note that the sum of the (ln prob) is the product of the prob
nwalker= 200
nchains= 1500
ndim = 4
fguess = [1,2,3,4] # dumb firt guesses
pos = [fguess + sigma_gb*np.random.randn(ndim) for i in range(nwalkers)]
sampler = emcee.EnsembleSampler(nwalkers, ndim, lnprob, args=(x, y, abs(yerr)))
sampler.run_mcmc(pos, nchains, progress=True)
I'm kind of new with this approach so I'm not sure if this is possible or if I should use another tool, also forget about the issues of subtracting point to a a discrete variable , I've already handle that, this it's just a dumb example to be more concise.
How I can make that the sampler just look around the values that I have a model?
