• ### Robust and Accurate Inference via a Mixture of Gaussian and Student's t Errors(1707.03057)

Aug. 17, 2018 stat.ME
A Gaussian measurement error assumption, i.e., an assumption that the data are observed up to Gaussian noise, can bias any parameter estimation in the presence of outliers. A heavy tailed error assumption based on Student's t distribution helps reduce the bias. However, it may be less efficient in estimating parameters if the heavy tailed assumption is uniformly applied to all of the data when most of them are normally observed. We propose a mixture error assumption that selectively converts Gaussian errors into Student's t errors according to latent outlier indicators, leveraging the best of the Gaussian and Student's t errors; a parameter estimation can be not only robust but also accurate. Using simulated hospital profiling data and astronomical time series of brightness data, we demonstrate the potential for the proposed mixture error assumption to estimate parameters accurately in the presence of outliers.
• ### How proper are Bayesian models in the astronomical literature?(1712.03549)

May 4, 2018 astro-ph.IM
The well-known Bayes theorem assumes that a posterior distribution is a probability distribution. However, the posterior distribution may no longer be a probability distribution if an improper prior distribution (non-probability measure) such as an unbounded uniform prior is used. Improper priors are often used in the astronomical literature to reflect a lack of prior knowledge, but checking whether the resulting posterior is a probability distribution is sometimes neglected. It turns out that 24 articles out of 75 articles (32%) published online in two renowned astronomy journals (ApJ and MNRAS) between Jan 1, 2017 and Oct 15, 2017 make use of Bayesian analyses without rigorously establishing posterior propriety. A disturbing aspect is that a Gibbs-type Markov chain Monte Carlo (MCMC) method can produce a seemingly reasonable posterior sample even when the posterior is not a probability distribution (Hobert and Casella, 1996). In such cases, researchers may erroneously make probabilistic inferences without noticing that the MCMC sample is from a non-existing probability distribution. We review why checking posterior propriety is fundamental in Bayesian analyses, and discuss how to set up scientifically motivated proper priors.
• ### The NANOGrav 11-year Data Set: High-precision timing of 45 Millisecond Pulsars(1801.01837)

April 10, 2018 astro-ph.IM, astro-ph.HE
We present high-precision timing data over time spans of up to 11 years for 45 millisecond pulsars observed as part of the North American Nanohertz Observatory for Gravitational Waves (NANOGrav) project, aimed at detecting and characterizing low-frequency gravitational waves. The pulsars were observed with the Arecibo Observatory and/or the Green Bank Telescope at frequencies ranging from 327 MHz to 2.3 GHz. Most pulsars were observed with approximately monthly cadence, with six high--timing-precision pulsars observed weekly, and all were observed at widely separated frequencies at each observing epoch in order to fit for time-variable dispersion delays. We describe our methods for data processing, time-of-arrival (TOA) calculation, and the implementation of a new, automated method for removing outlier TOAs. We fit a timing model for each pulsar that includes spin, astrometric, and, if necessary, binary parameters, in addition to time-variable dispersion delays and parameters that quantify pulse-profile evolution with frequency. The new timing solutions provide three new parallax measurements, two new Shapiro delay measurements, and two new measurements of large orbital-period variations. We fit models that characterize sources of noise for each pulsar. We find that 11 pulsars show significant red noise, with generally smaller spectral indices than typically measured for non-recycled pulsars, possibly suggesting a different origin. Future papers will use these data to constrain or detect the signatures of gravitational-wave signals.
• ### The Local Nanohertz Gravitational-Wave Landscape From Supermassive Black Hole Binaries(1708.03491)

Aug. 11, 2017 astro-ph.GA
Supermassive black hole binaries (SMBHBs) in the 10 million to 10 billion $M_\odot$ range form in galaxy mergers, and live in galactic nuclei with large and poorly constrained concentrations of gas and stars. There are currently no observations of merging SMBHBs--- it is in fact possible that they stall at their final parsec of separation and never merge. While LIGO has detected high frequency GWs, SMBHBs emit GWs in the nanohertz to millihertz band. This is inaccessible to ground-based interferometers, but possible with Pulsar Timing Arrays (PTAs). Using data from local galaxies in the 2 Micron All-Sky Survey, together with galaxy merger rates from Illustris, we find that there are on average $91\pm7$ sources emitting GWs in the PTA band, and $7\pm2$ binaries which will never merge. Local unresolved SMBHBs can contribute to GW background anisotropy at a level of $\sim20\%$, and if the GW background can be successfully isolated, GWs from at least one local SMBHB can be detected in 10 years.
• ### The NANOGrav Nine-year Data Set: Mass and Geometric Measurements of Binary Millisecond Pulsars(1603.00545)

Oct. 5, 2016 astro-ph.HE
We analyze 24 binary radio pulsars in the North American Nanohertz Observatory for Gravitational Waves (NANOGrav) nine-year data set. We make fourteen significant measurements of Shapiro delay, including new detections in four pulsar-binary systems (PSRs J0613$-$0200, J2017+0603, J2302+4442, and J2317+1439), and derive estimates of the binary-component masses and orbital inclination for these MSP-binary systems. We find a wide range of binary pulsar masses, with values as low as $m_{\rm p} = 1.18^{+0.10}_{-0.09}\text{M}_{\odot}$ for PSR J1918$-$0642 and as high as $m_{\rm p} = 1.928^{+0.017}_{-0.017}\text{M}_{\odot}$ for PSR J1614$-$2230 (both 68.3\% credibility). We make an improved measurement of the Shapiro timing delay in the PSR J1918$-$0642 and J2043+1711 systems, measuring the pulsar mass in the latter system to be $m_{\rm p} = 1.41^{+0.21}_{-0.18}\text{M}_{\odot}$ (68.3\% credibility) for the first time. We measure secular variations of one or more orbital elements in many systems, and use these measurements to further constrain our estimates of the pulsar and companion masses whenever possible. In particular, we used the observed Shapiro delay and periastron advance due to relativistic gravity in the PSR J1903+0327 system to derive a pulsar mass of $m_{\rm p} = 1.65^{+0.02}_{-0.02}\text{M}_{\odot}$ (68.3\% credibility). We discuss the implications that our mass measurements have on the overall neutron-star mass distribution, and on the "mass/orbital-period" correlation due to extended mass transfer.
• ### The NANOGrav Nine-year Data Set: Monitoring Interstellar Scattering Delays(1601.04490)

Jan. 18, 2016 astro-ph.IM, astro-ph.HE
We report on an effort to extract and monitor interstellar scintillation parameters in regular timing observations collected for the NANOGrav pulsar timing array. Scattering delays are measured by creating dynamic spectra for each pulsar and observing epoch of wide-band observations centered near 1500 MHz and carried out at the Green Bank Telescope and the Arecibo Observatory. The ~800-MHz wide frequency bands imply dramatic changes in scintillation bandwidth across the bandpass, and a stretching routine has been included to account for this scaling. For most of the 10 pulsars for which the scaling has been measured, the bandwidths scale with frequency less steeply than expected for a Kolmogorov medium. We find estimated scattering delay values that vary with time by up to an order of magnitude. The mean measured scattering delays are similar to previously published values and slightly higher than predicted by interstellar medium models. We investigate the possibility of increasing the timing precision by mitigating timing errors introduced by the scattering delays. For most of the pulsars, the uncertainty in the time of arrival of a single timing point is much larger than the maximum variation of the scattering delay, suggesting that diffractive scintillation remains only a negligible part of their noise budget.
• ### The NANOGrav Nine-year Data Set: Astrometric Measurements of 37 Millisecond Pulsars(1509.08982)

Jan. 2, 2016 astro-ph.GA, astro-ph.HE
Using the nine-year radio-pulsar timing data set from the North American Nanohertz Observatory for Gravitational Waves (NANOGrav), collected at Arecibo Observatory and the Green Bank Telescope, we have measured the positions, proper motions, and parallaxes for 37 millisecond pulsars. We report twelve significant parallax measurements and distance measurements, and eighteen lower limits on distance. We compare these measurements to distances predicted by the NE2001 interstellar electron density model and find them to be in general agreement. We use measured orbital-decay rates and spin-down rates to confirm two of the parallax distances and to place distance upper limits on other sources; these distance limits agree with the parallax distances with one exception, PSR J1024-0719, which we discuss at length. Using the proper motions of the 37 NANOGrav pulsars in combination with other published measurements, we calculate the velocity dispersion of the millisecond pulsar population in Galactocentric coordinates. We find the radial, azimuthal, and perpendicular dispersions to be 46, 40, and 24 km s-1, respectively, in a model that allows for high-velocity outliers; or 81, 58, and 62 km s-1 for the full population. These velocity dispersions are far smaller than those of the canonical pulsar population, and are similar to older Galactic disk populations. This suggests that millisecond pulsar velocities are largely attributable to their being an old population rather than being artifacts of their birth and evolution as neutron star binary systems. The components of these velocity dispersions follow similar proportions to other Galactic populations, suggesting that our results are not biased by selection effects.