arXiv subjects

Wenbin Lu

Publications and source records attributed to Wenbin Lu.

At least 109 records · Page 6Linked to original sources

arXiv2021

On Estimating Optimal Regime for Treatment Initiation Time Based on Restricted Mean Residual Lifetime

When to initiate treatment on patients is an important problem in many medical studies such as AIDS and cancer. In this article, we formulate the treatment initiation time problem for time-to-event data and propose an optimal individualized regime that determines the best treatment initiation time for individual patients based on their characteristics. Different from existing optimal treatment regimes where treatments are undertaken at a pre-specified time, here new challenges arise from the complicated missing mechanisms in treatment initiation time data and the continuous treatment rule in terms of initiation time. To tackle these challenges, we propose to use restricted mean residual lifetime as a value function to evaluate the performance of different treatment initiation regimes, and develop a nonparametric estimator for the value function, which is consistent even when treatment initiation times are not completely observable and their distribution is unknown. We also establish the asymptotic properties of the resulting estimator in the decision rule and its associated value function estimator. In particular, the asymptotic distribution of the estimated value function is nonstandard, which follows a weighted chi-squared distribution. The finite-sample performance of the proposed method is evaluated by simulation studies and is further illustrated with an application to a breast cancer data.

stat.ME↗

arXiv2021

Online Testing of Subgroup Treatment Effects Based on Value Difference

Online A/B testing plays a critical role in the high-tech industry to guide product development and accelerate innovation. It performs a null hypothesis statistical test to determine which variant is better. However, a typical A/B test presents two problems: (i) a fixed-horizon framework inflates the false-positive errors under continuous monitoring; (ii) the homogeneous effects assumption fails to identify a subgroup with a beneficial treatment effect. In this paper, we propose a sequential test for subgroup treatment effects based on value difference, named SUBTLE, to address these two problems simultaneously. The SUBTLE allows the experimenters to "peek" at the results during the experiment without harming the statistical guarantees. It assumes heterogeneous treatment effects and aims to test if some subgroup of the population will benefit from the investigative treatment. If the testing result indicates the existence of such a subgroup, a subgroup will be identified using a readily available estimated optimal treatment rule. We examine the empirical performance of our proposed test on both simulations and a real dataset. The results show that the SUBTLE has high detection power with controlled type I error at any time, is more robust to noise covariates, and can achieve early stopping compared with the corresponding fixed-horizon test.

stat.ME↗

arXiv2021

The nascent milliquasar VT J154843.06+220812.6: tidal disruption event or extreme accretion-state change?

We present detailed multiwavelength follow up of a nuclear radio flare, VT J154843.06+220812.6, hereafter VT J1548. VT J1548 was selected as a ${\sim}1$ mJy radio flare in 3 GHz observations from the VLA Sky Survey (VLASS). It is located in the nucleus of a low mass ($\log M_{\rm BH}/M_\odot \sim6$) host galaxy with weak or no past AGN activity. VT J1548 is associated with a slow rising (multiple year), bright mid IR flare in the WISE survey, peaking at ${\sim}10\%L_{\rm edd.}$. No associated optical transient is detected, although we cannot rule out a short, early optical flare given the limited data available. Constant late time (${\sim}3$ years post-flare) X-ray emission is detected at ${\sim}10^{42}$ erg s$^{-1}$. The radio SED is consistent with synchrotron emission from an outflow incident on an asymmetric medium. A follow-up, optical spectrum shows transient, bright, high-ionization coronal line emission ($[{\rm Fe\,X}]\,λ6375,[{\rm Fe\,XI}]\,λ7894,[{\rm S\,XII}]\,λ7612$). Transient broad H$α$ is also detected but without corresponding broad H$β$ emission, suggesting high nuclear extinction. We interpret this event as either a tidal disruption event or an extreme flare of an active galactic nucleus, in both cases obscured by a dusty torus. Although these individual properties have been observed in previous transients, the combination is unprecedented. This event highlights the importance of searches across all wave bands for assembling a sample of nuclear flares that spans the range of observable properties and possible triggers.

astro-ph.HE↗

arXiv2021

Infrared dust echoes from neutron star mergers

A significant fraction of binary neutron star mergers occur in star-forming galaxies where the UV-optical and soft X-ray afterglow emission from the relativistic jet may be absorbed by dust and re-emitted at longer wavelengths. We show that, for mergers occurring in gas-rich environment (n_H > 0.5 cm^{-3} at a few to tens of pc) and when the viewing angle is less than about 30 degrees, the emission from heated dust should be detectable by James Webb Space Telescope (JWST), with a detection rate of the order once per year. The spatial separation between the dust emission and the merger site is a few to 10 milli-arcsecs (for a source distance of 150 Mpc), which may be astrometrically resolved by JWST for sufficiently high signal-noise-ratio detections. Measuring the superluminal apparent speed of the flux centroid directly gives the orbital inclination of the merger, which can be combined with gravitational wave data to measure the Hubble constant. For a line of sight within the jet opening angle, the dust echoes are much brighter and may contaminate the search for kilonova candidates from short gamma-ray bursts, such as the case of GRB 130603B.

astro-ph.HE↗

arXiv2021

Kinematics of Crab Giant Pulses

The Crab Pulsar's radio emission is unusual, consisting predominantly of giant pulses, with durations of about a micro-second but structure down to the nano-second level, and brightness temperatures of up to $10^{37}\,$K. It is unclear how giant pulses are produced, but they likely originate near the pulsar's light cylinder, where corotating plasma approaches the speed of light. We report observations in the 400-800 MHz frequency band, where the pulses are broadened by scattering in the surrounding Crab nebula. We find that some pulse frequency spectra show strong bands, which vary during the scattering tail, in one case showing a smooth upward drift. While the banding may simply reflect interference between nano-second scale pulse components, the variation is surprising, as in the scattering tail the only difference is that the source is observed via slightly longer paths, bent by about an arcsecond in the nebula. The corresponding small change in viewing angle could nevertheless reproduce the observed drift by a change in Doppler shift, if the plasma that emitted the giant pulses moved highly relativistically, with a Lorentz factor $γ\sim10^4$ (and without much spread in $γ$). If so, this would support models that appeal to highly relativistic plasma to transform ambient magnetic structures to coherent GHz radio emission, be it for giant pulses or for potentially related sources, such as fast radio bursts.

astro-ph.HE↗

arXiv2021

Survival times of supramassive neutron stars resulting from binary neutron star mergers

A binary neutron star (BNS) merger can lead to various outcomes, from indefinitely stable neutron stars, through supramassive (SMNS) or hypermassive (HMNS) neutron stars supported only temporarily against gravity, to black holes formed promptly after the merger. Up-to-date constraints on the BNS total mass and the neutron star equation of state suggest that a long-lived SMNS may form in $\sim 0.45-0.9$ of BNS mergers. A maximally rotating SMNS needs to lose $\sim 3-6\times 10^{52}$ erg of it's rotational energy before it collapses, on a fraction of the spin-down timescale. A SMNS formation imprints on the electromagnetic counterparts to the BNS merger. However, a comparison with observations reveals tensions. First, the distribution of collapse times is too wide and that of released energies too narrow (and the energy itself too large) to explain the observed distributions of internal X-ray plateaus, invoked as evidence for SMNS-powered energy injection. Secondly, the immense energy injection into the blastwave should lead to extremely bright radio transients which previous studies found to be inconsistent with deep radio observations of short gamma-ray bursts. Furthermore, we show that upcoming all-sky radio surveys will constrain the extracted energy distribution, independently of a GRB jet formation. Our results can be self-consistently understood, provided that most BNS merger remnants collapse shortly after formation (even if their masses are low enough to allow for SMNS formation). This naturally occurs if the remnant retains half or less of its initial energy by the time it enters solid body rotation.

astro-ph.HE↗

arXiv2021

GEAR: On Optimal Decision Making with Auxiliary Data

Personalized optimal decision making, finding the optimal decision rule (ODR) based on individual characteristics, has attracted increasing attention recently in many fields, such as education, economics, and medicine. Current ODR methods usually require the primary outcome of interest in samples for assessing treatment effects, namely the experimental sample. However, in many studies, treatments may have a long-term effect, and as such the primary outcome of interest cannot be observed in the experimental sample due to the limited duration of experiments, which makes the estimation of ODR impossible. This paper is inspired to address this challenge by making use of an auxiliary sample to facilitate the estimation of ODR in the experimental sample. We propose an auGmented inverse propensity weighted Experimental and Auxiliary sample-based decision Rule (GEAR) by maximizing the augmented inverse propensity weighted value estimator over a class of decision rules using the experimental sample, with the primary outcome being imputed based on the auxiliary sample. The asymptotic properties of the proposed GEAR estimators and their associated value estimators are established. Simulation studies are conducted to demonstrate its empirical validity with a real AIDS application.

stat.ME↗

arXiv2021

Fast Optical Transients from Stellar-Mass Black Hole Tidal Disruption Events in Young Star Clusters

Observational evidence suggests that the majority of stars may have been born in stellar clusters or associations. Within these dense environments, dynamical interactions lead to high rates of close stellar encounters. A variety of recent observational and theoretical indications suggest stellar-mass black holes may be present and play an active dynamical role in stellar clusters of all masses. In this study, we explore the tidal disruption of main sequence stars by stellar-mass black holes in young star clusters. We compute a suite of over 3000 independent $N$-body simulations that cover a range in cluster mass, metallicity, and half-mass radii. We find stellar-mass black hole tidal disruption events (TDEs) occur at an overall rate of up to roughly $200\,\rm{Gpc}^{-3}\,\rm{yr}^{-1}$ in young stellar clusters in the local universe. These TDEs are expected to have several characteristic features, namely fast rise times of order a day, peak X-ray luminosities of at least $10^{44}\,\rm{erg\,s}^{-1}$, and bright optical luminosities (roughly $10^{41}-10^{44}\,\rm{erg\,s}^{-1}$) associated with reprocessing by a disk wind. In particular, we show these events share many features in common with the emerging class of Fast Blue Optical Transients.

astro-ph.HE↗

arXiv2021

First light from tidal disruption events

When a star comes too close to a supermassive black hole, it gets torn apart by strong tidal forces in a tidal disruption event, or TDE. Half of the elongated stream of debris comes back to the stellar pericenter where relativistic apsidal precession induces a self-crossing shock. As a result, the gas gets launched into an outflow that can experience additional interactions, leading to the formation of an accretion disc. We carry out the first radiation-hydrodynamics simulations of this process, making use of the same injection procedure to treat the self-crossing shock as in our previous adiabatic study (Bonnerot & Lu 2020). Two sets of realistic parameters of the problem are considered that correspond to different strengths of this initial interaction. In both cases, we find that the injected matter has its trajectories promptly circularized by secondary shocks taking place near the black hole. However, the generated internal energy efficiently diffuses away in the form of radiation, which results in a thin vertical profile of the formed disc. The diffusing photons promptly irradiate the surrounding debris until they emerge with a bolometric luminosity of $L\approx 10^{44} \, \rm erg\, s^{-1}$. Towards the self-crossing shock, diffusion is however slowed that results in a shallower luminosity increase, with a potentially significant component in the optical band. Matter launched to large distances continuously gains energy through radiation pressure, which can cause a significant fraction to become unbound. This work provides direct insight into the origin of the early emission from TDEs, which is accessed by a rapidly increasing number of observations.

astro-ph.HE↗

arXiv2021

The former companion of the hyper-velocity star S5-HVS1

The hyper-velocity star S5-HVS1, ejected 5 Myr ago from the Galactic Center at 1800 km/s, was most likely produced by tidal break-up of a tight binary by the supermassive black hole SgrA*. Taking a Monte Carlo approach, we show that the former companion of S5-HVS1 was likely a main-sequence star between 1.2 and 6 solar masses and was captured into a highly eccentric orbit with pericenter distance in the range 1-10 AU and semimajor axis about $10^3$ AU. We then explore the fate of the captured star. We find that the heat deposited by tidally excited stellar oscillation modes leads to runaway disruption if the pericenter distance is smaller than about 3 AU. Over the past 5 Myr, its angular momentum has been significantly modified by orbital relaxation, which may stochastically drive the pericenter inwards below 3 AU and cause tidal disruption. We find an overall survival probability in the range 5% to 50%, depending on the local relaxation time in the close environment of the captured star, and the initial pericenter at capture. The pericenter distance of the surviving star has migrated to 10-100 AU, making it potentially the most extreme member of the S-star cluster. From the ejection rate of S5-HVS1-like stars, we estimate that there may currently be a few stars in such highly eccentric orbits. They should be detectable (typically Ks < 18.5 mag) by the GRAVITY instrument and by future Extremely Large Telescopes and hence provide an extraordinary probe of the spin of SgrA*.

astro-ph.HE↗

arXiv2020

On the formation of GW190814

The LIGO-Virgo collaboration recently reported a puzzling event, GW190814, with component masses of 23 and 2.6 solar masses. Motivated by the relatively small rate of such a coalescence and the fact that the mass of the secondary is close to the total mass of known binary neutron star (bNS) systems, we propose that GW190814 was a second-generation merger from a hierarchical triple system, i.e., the remnant from the bNS coalescence was able to merge again with the 23 Msun black hole (BH) tertiary. We show that this occurs at a sufficiently high probability provided that the semimajor axis of the outer orbit is less than a few AU at the time of bNS coalescence. It remains to be explored whether the conditions for the formation of such tight triple systems are commonly realized in the Universe, especially in low metallicity (less than 0.1 solar) environments. Our model provides a number of predictions. (1) The spin of the secondary in GW190814-like systems is 0.6 to 0.7. (2) The component mass distribution from a large sample of LIGO sources should have a narrow peak between 2.5 and ~3.5 Msun, whereas the range between ~3.5 and ~5 Msun stays empty (provided that stellar evolution does not generate such BHs in the "mass gap"). (3) About 90% (10%) of GW190814-like events have an eccentricity of e > 2x10^{-3} (> 0.1) near gravitational wave frequency of 10 mHz. (4) A significant fraction (> 10%) of bNS mergers should have signatures of a massive tertiary at a distance of a few AU in the gravitational waveform. (5) There are 10^5 undetected radio-quiet bNS systems with a massive BH tertiary in the Milky Way.

astro-ph.HE↗

arXiv2020

Statistical Inference for Online Decision-Making: In a Contextual Bandit Setting

Online decision-making problem requires us to make a sequence of decisions based on incremental information. Common solutions often need to learn a reward model of different actions given the contextual information and then maximize the long-term reward. It is meaningful to know if the posited model is reasonable and how the model performs in the asymptotic sense. We study this problem under the setup of the contextual bandit framework with a linear reward model. The $\varepsilon$-greedy policy is adopted to address the classic exploration-and-exploitation dilemma. Using the martingale central limit theorem, we show that the online ordinary least squares estimator of model parameters is asymptotically normal. When the linear model is misspecified, we propose the online weighted least squares estimator using the inverse propensity score weighting and also establish its asymptotic normality. Based on the properties of the parameter estimators, we further show that the in-sample inverse propensity weighted value estimator is asymptotically normal. We illustrate our results using simulations and an application to a news article recommendation dataset from Yahoo!.

stat.ML↗

arXiv2020

Statistical Inference for Online Decision Making via Stochastic Gradient Descent

Online decision making aims to learn the optimal decision rule by making personalized decisions and updating the decision rule recursively. It has become easier than before with the help of big data, but new challenges also come along. Since the decision rule should be updated once per step, an offline update which uses all the historical data is inefficient in computation and storage. To this end, we propose a completely online algorithm that can make decisions and update the decision rule online via stochastic gradient descent. It is not only efficient but also supports all kinds of parametric reward models. Focusing on the statistical inference of online decision making, we establish the asymptotic normality of the parameter estimator produced by our algorithm and the online inverse probability weighted value estimator we used to estimate the optimal value. Online plugin estimators for the variance of the parameter and value estimators are also provided and shown to be consistent, so that interval estimation and hypothesis test are possible using our method. The proposed algorithm and theoretical results are tested by simulations and a real data application to news article recommendation.

stat.ML↗

arXiv2020

A unified picture of Galactic and cosmological fast radio bursts

The discovery of a fast radio burst (FRB) in our galaxy associated with a magnetar (neutron star with strong magnetic field) has provided a critical piece of information to help us finally understand these enigmatic transients. We show that the volumetric rate of Galactic-FRB like events is consistent with the faint end of the cosmological FRB rate, and hence they most likely belong to the same class of transients. The Galactic FRB had an accompanying X-ray burst but many X-ray bursts from the same object had no radio counterpart. Their relative rates suggest that for every FRB there are roughly 10^2 to 10^3 X-ray bursts. The radio lightcurve of the galactic FRB had two spikes separated by 30 ms in the 400-800 MHz frequency band. This is an important clue and highly constraining of the class of models where the radio emission is produced outside the light-cylinder of the magnetar. We suggest that magnetic disturbances close to the magnetar surface propagate to a distance of a few tens of neutron star radii where they damp and produce radio emission. The coincident hard X-ray spikes associated with the two FRB pulses seen in this burst and the flux ratio between the two frequency bands can be understood in this scenario. This model provides a unified picture for faint bursts like the Galactic FRB as well as the bright events seen at cosmological distances.

astro-ph.HE↗

arXiv2020

Imprint of local environment on fast radio burst observations

When fast radio burst (FRB) waves propagate through the local (< 1 pc) environment of the FRB source, electrons in the plasma undergo large-amplitude oscillations. The finite-amplitude effects cause the effective plasma frequency and cyclotron frequency to be dependent on the wave strength. The dispersion measure and rotation measure should therefore vary slightly from burst to burst for a repeating source, depending on the luminosity and frequency of the individual burst. Furthermore, free-free absorption of strong waves is suppressed due to the accelerated electrons' reduced energy exchange in Coulomb collisions. This allows bright low-frequency bursts to propagate through an environment that would be optically thick to low-amplitude waves. Given a large sample of bursts from a repeating source, it would be possible to use the deficit of low-frequency and low-luminosity bursts to infer the emission measure of the local intervening plasma and its distance from the source. Information about the local environment will shed light on the nature of FRB sources.

astro-ph.HE↗

arXiv2020

Implications of CHIME repeating fast radio bursts

CHIME has now detected 18 repeating fast radio bursts (FRBs). We explore what can be learned about the energy distribution and activity level of the repeaters by fitting realistic FRB population models to the data. For a power-law energy distribution dN/dE ~ E^{-γ} for the repeating bursts, there is a critical index γ_c that controls whether the dispersion measure (DM, a proxy for source distance) distribution of repeaters is bottom or top-heavy. We find γ_c = 7/4 for Poisson wait-time distribution of repeaters in Euclidean space and further demonstrate how it is affected by temporal clustering of repetitions and cosmological effects. It is especially interesting that two of the CHIME repeaters (FRB 181017 and 190417) have large DM ~ 1000 pc/cm^3. These can be understood if: (i) the energy distribution is shallow γ= 1.7^{+0.3}_{-0.1} (68% confidence) or (ii) a small fraction of sources are extremely active. In the second scenario, these two high-DM sources should be repeating more than 100 times more frequently than FRB 121102, and the energy index is constrained to be γ= 1.9^{+0.3}_{-0.2} (68% confidence). In either case, this γis consistent with the energy dependence of the non-repeating ASKAP sample, which suggests that they are drawn from the same population. Finally, our model predicts how the CHIME repeating fraction should decrease with redshift, and this can be compared with observations to infer the distribution of activity level in the whole population.

astro-ph.HE↗

arXiv2020

Causal Effect Estimation and Optimal Dose Suggestions in Mobile Health

In this article, we propose novel structural nested models to estimate causal effects of continuous treatments based on mobile health data. To find the treatment regime that optimizes the expected short-term outcomes for patients, we define a weighted lag-K advantage as the value function. The optimal treatment regime is then defined to be the one that maximizes the value function. Our method imposes minimal assumptions on the data generating process. Statistical inference is provided for the estimated parameters. Simulation studies and an application to the Ohio type 1 diabetes dataset show that our method could provide meaningful insights for dose suggestions with mobile health data.

stat.ME↗

arXiv2020

Kernel Assisted Learning for Personalized Dose Finding

An individualized dose rule recommends a dose level within a continuous safe dose range based on patient level information such as physical conditions, genetic factors and medication histories. Traditionally, personalized dose finding process requires repeating clinical visits of the patient and frequent adjustments of the dosage. Thus the patient is constantly exposed to the risk of underdosing and overdosing during the process. Statistical methods for finding an optimal individualized dose rule can lower the costs and risks for patients. In this article, we propose a kernel assisted learning method for estimating the optimal individualized dose rule. The proposed methodology can also be applied to all other continuous decision-making problems. Advantages of the proposed method include robustness to model misspecification and capability of providing statistical inference for the estimated parameters. In the simulation studies, we show that this method is capable of identifying the optimal individualized dose rule and produces favorable expected outcomes in the population. Finally, we illustrate our approach using data from a warfarin dosing study for thrombosis patients.

stat.ME↗