About this Research Topic
Astrostatistics is developing fast as astronomy is confronted with huge volumes of data. This constitutes undoubtedly a new era through a move from an object- to data-driven astrophysics. We can distinguish two types of challenges: the first one is related to the exploration of the data space and the second one is the physical interpretation of the results of this exploration. Astrostatistics cannot be reduced to the technical issues brought by methods and algorithms required to analyze the data. Despite the fact that these tools are not part of the curriculum of astronomers, things are changing rapidly through schools, training sessions, collaborations with experts (statisticians) as well as developments of specific tools adapted to astronomical data. The use of new techniques pervades the astronomical literature. To convince more and more astronomers to switch to this new approach in science, it is important to demonstrate the relevance of astrostatistics results. This is the second type of challenge astrostatistics faces.
The question that needs to be solved is the following: how can we reconnect statistical results obtained by automatic tools such as machine learning or hypothesis testing, with object-driven inferences derived from visual inspection, modelization or numerical simulation? How do we reconcile a statistical, hence fuzzy, classification with the visual Hubble sequence of galaxies? How do we describe, synthetize, the data cubes (images + spectra) obtained by IFUs? How can we constrain a model of stellar evolution with the statistics of billions of stars? How do we characterize a scaling law in a 4-D or 10-D dimension parameter space?
Some of the specific topics that could be discussed in this Research Topic are the interpretability of machine learning and statistical results; the use of unsupervised clustering of data, spectra, and images; the selection of variables depending on their availability; the fact that correlation is not causality (what are the latent variables?); the outlier detection and characterization in the context of Big Data; the hypothesis testing techniques to evaluate the confidence in detection of faint sources; fuzzy classification and probabilistic physics; astrostatistics in numerical simulations; etc.
We want to focus particularly on the connection between the tools and the astrophysics, between data analysis and science. Similarly to telescopes and detectors that observe the Universe, astrostatistics is an ensemble of techniques that observe the data: what do we see and how can we understand?
Keywords: Astrophysics, Statistics, Data Analysis, Machine Learning, Physical Interpretation
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.