Astrostatistics as a New Statistical Discipline



Download 11.53 Kb.
Date20.05.2017
Size11.53 Kb.
#18675
Astrostatistics as a New Statistical Discipline
Joseph M. Hilbe

President, International Astrostatistics Association


Astronomers have had an interesting relationship with statistics over the years. Early Greek natural philosophers used the mean, median, range and other statistics to calculate the position of various celestial objects. Hipparchus (190-120BCE) calculated the distances from the Earth to the Sun and Moon in terms of Earth radii. He concluded that the median distance from the Earth to the Moon is 60.5 Earth radii. In fact, it is 60.3 radii, an incredibly close estimate given the tools he had to work with.

Astronomers have continued to employ descriptive methods to astronomical data throughout the years, and were the first to use linear regression in a scientific study. In 1801 a Hungarian astronomer named Franz von Zach used the new inferential statistical method developed by Carl Gauss a few years before in order to predict the position of the asteroid Ceres as it came to view from behind the Sun. However, astronomers soon began to turn to calculus and differential equations rather than to statistics when analyzing astronomical data. They also focused on instrumentation such as the spectroscope and more powerful telescopes. Except for isolated exceptions, astronomers have largely ignored the advances made in statistics since the early nineteenth century, thinking that the field of statistics had little to offer astronomy.

This attitude began to change during the final two decades of the last millennium. The reason, I believe, is due to the advances made in computing power, and especially due to the introduction of the personal computer in August 1981. During the 1980s and 1990s we saw a truly amazing increase in computing speed, RAM and storage capability. During this period statistical capabilities progressed equally fast, taking advantage of the increasing developments in computer technology. A number of statistical packages emerged during this period, each competing for market share, which meant developing fast and efficient statistical procedures that earlier statisticians could only dream about. By the beginning of the new millennium computing power became sophisticated enough to allow development of MCMC-based Bayesian modeling. Sampling algorithms became ever more efficient and fast during the first decade of the 2000s, which made many astronomers take notice. Astronomers could finally see how they could use – and how they needed – statistics to better understand their data.

Actually, because of enhanced statistical capabilities and computing power, a few conferences began to he held about using statistical methods for analyzing astronomical data. The first was perhaps the “Statistical Challenges in Modern Astronomy” (SCMA) conferences started by astronomer Eric Feigelson and statistician Jogesh Babu of Pennsylvania State University in 1991. These conferences have been held at five year intervals since then, drawing the interest of a hundred or more researchers at each conference. Likewise, in 1991 the first of the annual “Astronomical Data Analysis Software and Systems” (ADASS) conferences began. Several rather isolated collaborations of astronomers and statisticians began as well during this period. The SCMA conferences were quite successful, but generally did not bring many new researchers into the area. ADASS was not involved with statistical analysis, but rather focused on what we now call astroinformatics. Astroinformatics relates to the gathering, storage, and manipulation of astronomical data. Since statisticians need data, astroinformatics is vitally important to the success of astrostatistics. But it is not, strictly speaking, statistics.

By 2009 the general complacency, or perhaps skepticism, which the majority of astrophysicists felt toward statistical modeling over the past century and a half began to radically change. More astronomers could finally see that statistical modeling can be of substantial use to them in understanding their data. The time was right to join astronomers and statisticians together from around the world in the creation of a new scientific discipline, a discipline in which astronomers and statisticians could collaborate to produce more accurate results. It was also important to have the foremost statistical and astronomical societies recognize astrostatistics as a viable discipline under their joint purview. At the beginning of 2008 there were no astrostatistics committees, or even an interest group that was authorized within any statistical or astronomical association. If the science was to truly advance this was necessary, at least to start with.

In 2008 I formed an astrostatistics interest group within the International Statistical Institute (ISI). It was the first such interest group or committee ever authorized under an astronomical or statistical association. The group met at the 2009 World Statistics Congress in Durban, South Africa, and proposed becoming a standing committee within the ISI. The ISI executive board approved it in December of that year. By January we had far more members than allowed for an ISI committee. We then became the ISI International Astrostatistics Network.

The Network continued to grow until 2012 when our executive board approved my motion to become an independent International Astrostatistics Association (IAA). Also in 2012 Feigelson and I developed the Astrostatistics and Astroinformatics Portal (ASAIP) (https://asaip.psu.edu), sponsored by the Penn State Department of Astronomy. The Portal provides the astrostatistics community with free access to a host of publications, and information regarding all forthcoming conferences, workshops and tutorials related to astrostatistics. The Portal was a key component to the development of astrostatistics as a new scientific discipline.

In 2014 the International Astronomical Union (IAU) approved the first ever IAU symposium on astrostatistics. Held in Lisbon, Portugal on the statistics of cosmology, the symposium was an outstanding success. A second Symposium has been approved for 2017, while the first IAU symposium on astroinformatics will occur later this year in Italy. Equally important, the IAU approved the first Commission on Astroinformatics and Astrostatistics, which became effective following the 2015 IAU General Assembly in Honolulu, with Feigelson as its initial President. At the 2015 General Assembly the first Focus Meeting on an astrostatistical subject was held – on statistics and exoplanets. The Symposia and Focus Meeting were all co-sponsored by the IAA. Without a doubt the scientific community now has a new discipline in its fold – astrostatistics (in the broad sense of incorporating astroinformatics).



The IAA currently has nearly 600 members from 56 nations. New sections and committees have been formed and are working within the IAA. The IAA recently moved their offices to the Brera Astronomical Observatory in Milan, Italy. The IAA also approved a new permanent logo, established a new website (http://iaa.mi.oa-brera.inaf.it), and presented the association’s first awards. Awards presented were the “Outstanding Contributions to Astrostatistics” medal, our top award, “Elected IAA Fellow”, and three categories of “Outstanding Publication in Astrostatistics”. Cambridge University Press co-sponsored the awards.

Since the creation of the ISI astrostatistics committee and network in 2009, and then the IAA and ASAIP in 2012, astrostatistics has evolved from a few astronomers and even less statisticians collaborating on specific topics, and conferences being held each 5 years, to the point we are at today ---- a full-fledged scientific discipline with a viable international association and an IAU Commission. The ASAIP has close to a thousand members accessing astrostatistical and astroinformatical related publications and conferences, and we even have an IAA Facebook page which currently has 931 followers and is seen by an average of 2160 researchers per week. The IAA is attempting to create a discipline within which the astronomical and statistical communities, and those in the information sciences, can work together with the aim of making significant discoveries about the Universe. Recently the discovery of gravitational waves was made based on complex Bayesian modeling. Stay tuned for many more astronomical and physical discoveries, each having an important statistical component.

Download 11.53 Kb.

Share with your friends:




The database is protected by copyright ©ininet.org 2024
send message

    Main page