import pymc as mc
import models
# fit with MCMC, save results in a database db can be one of the
# following: no_trace, ram, pickle, txt, sqlite, mysql, hdf5
vars = models.nonlinear()
mcmc = mc.MCMC(vars, db='sqlite', dbname='nonlinear.sqlite')
mcmc.use_step_method(mc.AdaptiveMetropolis, [mcmc.beta, mcmc.gamma])
mcmc.sample(5000, 2500, 5)
# load the database from disk
db = mc.database.sqlite.load('nonlinear.sqlite')
print db.beta.stats()
# do it again with txt to compare size of output
vars = models.nonlinear()
mcmc = mc.MCMC(vars, db='txt', dbname='nonlinear.txt')
mcmc.use_step_method(mc.AdaptiveMetropolis, [mcmc.beta, mcmc.gamma])
mcmc.sample(5000, 2500, 5)
# load the database from disk
db = mc.database.txt.load('nonlinear.txt')
print db.beta.stats()
import pylab as pl
import pymc as mc
import models
import graphics
# make model
vars = models.nonlinear()
#vars['beta'].value = [10, -9, 15] # carefully choosen initial value, for demonstration purposes only
m = mc.MCMC(vars)
m.sample(iter=20000, burn=10000, thin=10)
# display results
pl.figure(figsize=(12,9))
graphics.plot_2005_data()
graphics.plot_nonlinear_model(m)
pl.savefig('../tex/ex2.png')
def test_nonlinear_model(self):
vars = models.nonlinear()
assert 'gamma' in vars
