Monday, Aug 5: 2:00 PM - 3:50 PM

1528

Topic-Contributed Paper Session

1528

Topic-Contributed Paper Session

Oregon Convention Center

Room: CC-D139

The data generated from large astronomical surveys and cutting-edge experiments in particle physics have revealed the vital role of statistics in conducting reliable and reproducible statistical analyses, i.e., minimizing the risk of false discoveries while maximizing the power of the detection tools adopted. In this setting, a crucial stage of the analysis of astrophysical data is that of assessing the presence of new signals. Given the complexity and the lack of regularity of the models under study, however, classical statistical tools are typically not applicable. As a result, by developing robust statistical solutions to address this problem, statisticians have the unprecedented opportunity to contribute to groundbreaking discoveries in physics and astronomy. While focusing on the novel data-driven frontiers of the physical sciences, this session aims to (i) stimulate the interest of the audience by showing how important astrophysical questions often translate into fundamental statistical questions, (ii) describe some of the most recent solutions proposed in the literature, and (iii) motivate new research developments in the intersection between statistics and the physical sciences while overcoming cross-disciplinary language barriers.

Yes

Section on Physical and Engineering Sciences

Astrostatistics Interest Group

Section on Statistical Computing

Just as our own Sun emits solar flares, other stars emit stellar flares. Astrophysicists are interested in the number and energy distribution of stellar flares because this information can (1) test theories about stellar magnetic fields, and (2) be used to predict the habitability of the star's planetary system. Stellar flares appear in the time series data of a star's brightness as a quick rise followed by an exponential decay back to the star's approximately stationary state. The current methods to detect stellar flares rely on "3-sigma clipping": a Gaussian process (GP) models the stationary state of the time series, and points beyond three standard deviations are identified as flares. The disadvantage of this approach is that the dimmer (and therefore less energetic) flares will be missed. In this talk, I will describe our new method for discovering stellar flares in time series data of M-dwarf stars. Our novel Bayesian approach combines a 3-state Hidden Markov Model with a GP, and identifies flares through state-decoding. Simulations show this method finds the smaller flares missed by 3-sigma clipping, and gives a more accurate flare energy distribution estimate.

This talk introduces a novel inferential strategy to test cosmological and astrophysical models for angular power spectra characterized by unknown parameters. We show that it is possible to assess the validity of such models without specifying the distribution of the estimators of the angular power spectrum being used. This holds true under rather general conditions, ensuring the applicability of the method across diverse scenarios. Moreover, the proposed solution overcomes the need for case-by-case simulations when testing different models - leading to remarkable advantages from a computational perspective.

New physics searches are usually done by training a supervised classifier to separate a signal model from a background model (known physics Standard Model). However, even when the signal model is correct, systematic errors in the background model can influence supervised classifiers and might adversely affect the signal detection procedure. To tackle this problem, one approach is to find a classifier constrained to be decorrelated with one or more protected variables, e.g. the invariant mass. Then use this classifier to find a signal-rich region where one can perform the signal detection test using the protected variable. We perform the decorrelation by considering an optimal transport map of the classifier output that makes it independent of the invariant mass for the background. We then estimate the signal strength in the signal-rich region to detect the presence of signal, using the experimental data on the invariant mass. We compare and contrast this decorrelation method with previous approaches, show that the decorrelation procedure is robust to background misspecification, and analyze the power of the test.

The challenge of predicting solar flares has been tackled frequently in recent years, thanks to the burst of high-quality solar data. Among all sources of solar data, the Helioseismic and Magnetic Imaging (HMI) and Atmospheric Imaging Assembly(AIA) data contain abundant information about the Sun in the format of multi-channel images, and applications of this data in deep learning models have been proven to be successful. However, these methods are not interpretable and cannot provide uncertainty quantification (UQ). In this talk, we develop an interpretable and trustworthy statistical method that treats the HMI/AIA data as tensor data. We propose to predict flare intensity with the tensor data in a novel framework called Tensor-GPST, where we first transform the high-dimensional tensor data into a low-dimensional latent tensor via sparse tensor contraction, and then the latent tensor is used for prediction via the Gaussian process. We introduce an anisotropic total-variation regularization when contracting the tensor and estimate the model with alternating proximal gradient descent. We validate our approach via simulation and real application to flare intensity forecasting.