Specfit

Infer polynomial coefficients and their covariance structure for fitting radio-astronometric callibrator spectra.

Author: Tim Molteno. tim@elec.ac.nz
Publication: Molteno, Timothy CA. "Correlation structure in flux-density calibrator models." Monthly Notices of the Royal Astronomical Society 527.3 (2024): 5732-5740.

Data

Machine readable data from this catalogue is available in the HDF file in the data directory. calibrator_catalogue.hdf5

Install

sudo pip3 install specfit

develop

pip3 install -e .

Examples

Here is an example. This code is in the examples directory.

import numpy as np
import specfit as sf
import matplotlib.pyplot as plt

# Data from J.E. Reynolds for J1939-6342
original_data = np.array(
    [[0.408,  6.24, 0.312 ],
     [0.843, 13.65, 0.6825],
     [1.38 , 14.96, 0.748 ],
     [1.413, 14.87, 0.7435],
     [1.612, 14.47, 0.7235],
     [1.66 , 14.06, 0.703 ],
     [1.665, 14.21, 0.7105],
     [2.295, 11.95, 0.5975],
     [2.378, 11.75, 0.5875],
     [4.8  ,  5.81, 0.2905],
     [4.8  ,  5.76, 0.288 ],
     [4.835,  5.72, 0.286 ],
     [4.85 ,  5.74, 0.287 ],
     [8.415,  2.99, 0.1495],
     [8.42 ,  2.97, 0.1485],
     [8.64 ,  2.81, 0.1405],
     [8.64 ,  2.81, 0.1405]])

freq_ghz, mu, sigma = original_data.T
freq = freq_ghz*1e9

names, stats, a_cov, a_corr, idata = \
    sf.spectral_inference("J1939-6342", 
        freq=nu, mu=data, sigma=sigma, order=4, nu0=1.4e9)

Now we can plot the data and show the results.

fig, ax = sf.dataplot(plt, "J1939-6342", freq=freq, mu=data, sigma=sigma)

a = stats[0] # Means

nu = np.linspace(min_freq, max_freq, 100)
S = sf.flux(nu, a, nu0=1.4e9)
ax.plot(nu/1e9, S, label="polynomial fit")
ax.legend()
fig.tight_layout()
plt.show()

print(names, stats)
print(a_cov)

TODO

Incorporate some ideas on using variances of parameters and constraints on flux uncertainties in place of requiring an explicit assumption of the sigma (in the case of data-free inference)
Use smoothness as a prior (rather than model-order).

Changelog

0.4.0b1 Update the marginal_likelihood method to correctly no longer use inferenceData objects to avoid a bug in pymc. Return the relative marginal likelihood (rather than the log marginal likelihood)
0.3.0b3 Clean up to use the natural log throughout! (IMPORTANT) Use consistent way to get names of variables from the posterior Add machine readable hdf5 file output.
0.3.0b2 Use pymc and upgrade to newer versions.
0.2.0b4 Include a separate function (marginal_likelihood) for estimating the marginal likelihood using SMC Change the likelihood to use a Student's t distribution for robustness.
0.2.0b3 Fix examples, move to github automation for release information.
0.1.0b3 First functioning release.
0.1.0b4 [In progress] Add the frequency range to the full_column output. Return the inference data to allow further processing Improved plotting and postprocessing. Added posterior PDF helper plotting function (slow) Use different tuning depending on polynomial order Output to a file, including lists of alternate names

tmolteno / specfit