Astrostatistics: Intensive Research Session on Stellar Evolution – February 20-23, 2006

A basic tool for understanding stellar evolution from an observational point of view is the so-called/ color-magnitude diagram/ (CMD), which locates a star by its color on the ordinate and its magnitude (log luminosity) on the abscissa. If one observes a/ cluster/ of stars, all of which are at almost the same distance, one can plot/ observed/ magnitude on the abscissa instead of/ absolute/ (intrinsic) magnitude. Since all of the stars in the cluster can be presumed to be of the same age, the CMDs of clusters of varying ages we can provide information about how stars evolve. Significant improvements can be expected by employing modern statistical methods to analyze the data.

Mini-workshop Goals

  • Improving sampling – Initial work is with a model that has been implemented as an MCMC sampler. Experience has shown that getting good sampling and convergence to the stationary distribution is not easy, due to strong nonlinear correlations between the age of the cluster and the individual masses of the stars in the cluster. We hope to improve sampling significantly.
  • Model selection – It is estimated that about 50% of all stars are double. Both the color and magnitude of a double star are shifted on the CMD by a small amount. Therefore, each star in the cluster has to be modeled with some probability that it is multiple. This poses a model selection problem. A second model selection arises from the possibility that a star is a field star. We aim to solve these model selection problems.
  • Selection models – The faintest stars in a cluster may not be visible at all if the cluster is far enough away. This leads to a truncation problem, since many of these stars will be faint white dwarfs, and white dwarfs lend a significant amount of weight to the age determination of the cluster. We will work on methods to account for this truncation properly, so as not to bias the inferred age of the cluster.

Scheduled Participants

  • William Jefferys, (Universities of Texas and Vermont), Leader
  • Ted von Hippel, (Univ of Texas)
  • Steve DeGennaro, (Univ of Texas)
  • Elizabeth Jeffery, (Univ of Texas)
  • Nathan Stein, (Univ of Texas)
  • David van Dyk, (Univ of California, Irvine)
  • Tom Loredo, (Cornell)
  • Theodore Arthur, Sande (MIT)