jvines / BayesianGLS

An implementation of the Bayesian Generalized Lomb Scargle periodogram as described by Mortier et al. 2015

BayesianGLS

An implementation of the Bayesian Generalized Lomb Scargle periodogram as described by Mortier et al. 2015

Installation

Just clone the repository and then install using setup.py

$ cd BayesianGLS
$ python setup.py install

Usage

To use, simply import bayesian_generalized_lomb_scargle from bgls

import scipy as sp
from scipy.stats import norm, uniform
import numpy as np

from bgls import bayesian_generalized_lomb_scargle

np.random.seed(42)
a = 5
p = 5.67
f = 1 / p
errs = abs(norm(loc=0, scale=1).rvs(size=250))
t = uniform.rvs(size=250, loc=2, scale=100)
t.sort()
fun = a * sp.sin(2 * sp.pi * f * t) + norm.rvs(size=250, scale=1)

PP = sp.linspace(sp.diff(t).mean(), 2 * 26, 100000)
FF = 1 / PP
lnp = bayesian_generalized_lomb_scargle(fun, t, errs, FF)

plt.errorbar(x=t, y=fun, yerr=errs, marker='o', ls='')
plt.xlabel('time')
plt.ylabel('rv')
plt.savefig('synthetic_data.pdf', bbox_inches='tight')
plt.clf()

plt.semilogx(PP, lnp)
period = PP[sp.argmax(lnp)]
lab = '{:.4f}'.format(period)
plt.axvline(PP[sp.argmax(lnp)], lw=5, alpha=.3, c='g', label=lab)
plt.legend(loc=0)
plt.title('BGLS')
plt.xlabel('period')
plt.ylabel('normalized loglikelihood')
plt.savefig('BGLS.pdf', bbox_inches='tight')
plt.clf()

The results look like this:

Uninstallation

If you wish to remove BGLS, just go to the terminal and type pip uninstall bgls

This code is based of Mortier et al. 2015