ASAIP Editors' Blog, August 2013
* Stellar Initial Mass Function
Thousands of papers have been written in the past half-century on the shape of the mass distribution of stars; it is thought to give a crucial clue into the astrophysical of star formation. Classic statistical models of the Initial Mass Function include Salpeter's simple power law (300 citations/yr), Kroupa's broken power law (200 citations/yr) and Chabrier's lognormal plus a power law (250 citations/yr). New papers recommend the log-logistic function and the truncated beta distribution. An earlier analysis fit the stellar IMF to a double Pareto-lognormal distribution that naturally arises from a stochastic process developed by statistician William Reed. The debate goes on …
* Weak gravitational lensing
A community of observational cosmologists have participated in a competition to recover slight distortions of faint galaxies from realistic simulated fields subject to Dark Matter gravitational lensing. Thirty methods from nine teams competed in the GRavitaitonal lEnsing Accuracy Testing 2010 (GREAT10) Star Challenge. Results are reported here with some methodological details here. Local interpolators performed best, such as Inverse Distance Weighting (commonly used in Earth sciences) and radial basis functions. In related work, a model-based maximum likelihood procedure for galaxy shear measurement was released, as was a computationally efficient classification-based nonparametric method.
* Image reconstruction
For many decades, a favorite tool for restoring astronomical images, when the telescope point spread function is known, has been the maximum likelihood Lucy-Richardson algorithm. It was an early implementation of the EM Algorithm very widely used for MLE calculations in statistics. Recently, two statisticians improve on EM-type algorithms for image reconstruction with the introduction of prior constraints and Bayesian analysis. Both procedures, described here and here, look promising for astronomical use.
* Astrostatistical software packages
This summer witnessed the release of several new statistically oriented software products of interest to astronomers. PYMORPH brings the well-established GALFIT model for the decomposition of galaxy structure from a CCD image into Python with parallel CPU implementation compatible with Virtual Observatory protocols. CASOS (a rare astronomical paper published in a statistics journal) implements a new algorithm for detecting anomalies in high-dimensional databases based on subspace projections. The efficient libsharp library for computing spherical harmonic transforms for cosmic microwave background studies is released. NIFTy is a Python package calculates that provides mathematical operations on a wide range of n-dimensional signals. For radio astronomy, the Pulsar Evaluation Algorithm for Candidate Extraction (PEACE) was found to be efficient in identifying pulsar signals. A flexible Bayesian Inference Engine was released, designed for fitting high-dimensional nonlinear astrophysical models with complex posteriors. Finally, leaders in astroinformatics recommend that all astronomy papers publicly share their codes.
We have opened up a new Resource page on visualization of astronomical and astrophysical datasets. Some new free software products are linked, scientific visualization galleries are presented, and color schemes are discussed. A number of recent articles on astronomical visualization are linked. Additions are very welcome; contact the Editors.
* Recent books
Three multiauthor volumes on astrostatistics appeared in the last year, two from the new Springer's Series in Astrostatistics: lectures from a 2011 workshop in La Palma ES, from the 2011 ISI World Statistics Congress in Dublin IE, and a collection of reviews on machine learning in astronomy. Two graduate texts also emerged -- the second edition of Wall & Jenkins Practical Statisitcs for Astronomers, and the first edition of Feigelson & Babu's Modern Statistical Methods for Astronomy with R Applications. The latter won a prize for best astronomy book of 2012. However, the Editors need advice for entries on the page Selected books in informatics … let us know your favorites.
* Recent and forthcoming conferences
This summer, continuing a trend dating back several years, astrostatistics was well-represented in large meetings attended by thousands of statisticians. The Joint Statistical Meeting in Montreal CA had a session with six contributed talks. Larry Wasserman, a very experienced astrostatistician, presented the Rietz Lecture. There were five invited papers on astrostatistics and a Special Topics session on Source and Feature Detection in Astronomy at the 59th ISI World Statistics Congress in Hong Kong CN. The first official astrostatistics conference of the International Astronomical Union, IAU Symp 306 Statistical Challenges in 21st Century Cosmology, has been approved for May 2014 in Lisbon PT. Please consider attending this premier conference, and be sure to look ahead at other scheduled conferences here. The Editors very much welcome reviews of recent meetings from participants or organizers, either briefly for this blog or at greater length for an ASAIP Article.
Eric Feigelson & Joseph Hilbe