from __future__ import print_function
import numpy as np
from scipy import stats
import statsmodels.api as sm
from statsmodels.base.model import GenericLikelihoodModel
data = sm.datasets.spector.load()
data.exog = sm.add_constant(data.exog, prepend=False)
# in this dir
probit_mod = sm.Probit(data.endog, data.exog)
probit_res = probit_mod.fit()
loglike = probit_mod.loglike
score = probit_mod.score
mod = GenericLikelihoodModel(data.endog, data.exog*2, loglike, score)
res = mod.fit(method="nm", maxiter = 500)
def probitloglike(params, endog, exog):
"""
Log likelihood for the probit
"""
q = 2*endog - 1
X = exog
return np.add.reduce(stats.norm.logcdf(q*np.dot(X,params)))
mod = GenericLikelihoodModel(data.endog, data.exog, loglike=probitloglike)
res = mod.fit(method="nm", fargs=(data.endog,data.exog), maxiter=500)
print(res)
def optimize_IcIsIr(dt,
guessParams=[-1, -1, -1],
deadtime=1.e-5,
method='nm',
prior=[np.nan] * 3,
prior_sig=[np.nan] * 3,
**kwargs):
"""
This function optimizes the loglikelihood for the bin-free SSD analysis.
It returns the model fit for the most likely I1, I2, Ir.
INPUTS:
dt - Float, [seconds], list of photon inter-arrival times
params - Floats, [1/seconds], list of initial guess for I1, I2, Ir
if the guess is <0 then equally distribute the total flux amongst all params with a guess <0
deadtime - float, [seconds], MKID deadtime after photon event
method - optimization method. 'nm'=Nelder-Mead, 'ncg'=Newton-conjugate gradient, etc...
paramsFixed - booleans, fix param_i during optimization to the initial guess
**kwargs - additional kwargs to GenericLikelihoodModel.fit()
www.statsmodels.org/stable/dev/generated/statsmodels.base.model.LikelihoodModel.fit.html
OUTPUTS:
GenericLikelihoodModelResults Object
See www.statsmodels.org/stable/dev/generated/statsmodels.base.model.GenericLikelihoodModelResults.html
"""
if prior is None:
prior = [np.nan] * 3
prior = np.append(prior, [np.nan] * (3 - len(prior)))
if prior_sig is None:
prior_sig = [np.nan] * 3
prior_sig = np.append(prior_sig, [np.nan] * (3 - len(prior_sig)))
#Provide a reasonable guess everywhere that guessParams<0
#If a prior is given then make that the guess
#equally distributes the average flux amongst params<0
#ie. guess=[-1, 30, -1] and I_avg=330 then guess-->[150,30,150].
guessParams = np.asarray(guessParams)
assert len(
guessParams
) == 3, "Must provide a guess for I1, I2, Ir. Choose -1 for automatic guess."
if np.any(prior == None):
prior[prior == None] = np.nan
guessParams[np.isfinite(prior)] = prior[np.isfinite(prior)]
if np.any(guessParams < 0):
I_avg = (len(dt)) / np.sum(dt)
I_guess = (I_avg - np.sum(guessParams[guessParams >= 0])) / np.sum(
guessParams < 0)
guessParams[guessParams < 0] = max(I_guess, 0)
#Define some functions
loglike = partial(posterior,
dt=dt,
deadtime=deadtime,
prior=prior,
prior_sig=prior_sig)
score = partial(posterior_jacobian,
dt=dt,
deadtime=deadtime,
prior=prior,
prior_sig=prior_sig)
hess = partial(posterior_hessian,
dt=dt,
deadtime=deadtime,
prior=prior,
prior_sig=prior_sig)
#Setup model
endog = np.asarray(dt, dtype=[('dt', 'float64')])
names = np.asarray(['Ic', 'Is', 'Ir'])
exog = np.ones(len(endog),
dtype={
'names': names,
'formats': ['float64'] * len(names)
})
model = GenericLikelihoodModel(endog,
exog=exog,
loglike=loglike,
score=score,
hessian=hess)
try:
kwargs['disp']
except KeyError:
kwargs['disp'] = False #change default disp kwarg to false
#fit model
return model.fit(guessParams, method=method, **kwargs)