Previous | Session 153 | Next
K. A. Huyser, K. H. Knuth (NASA Ames), A. R. Hajian (USNO)
In the course of understanding data acquired from astronomical observations, astronomers often construct a model of the object based on both the known physics and acquired data. The construction of a model can become quite complicated in terms of both the description of the object and the amount of data for which the model must account. The latter is becoming a more pressing issue as astronomers are faced with increasingly large datasets.
The technique of Bayesian modeling proceeds as follows. Given a parametric model of the object, one generates a synthetic dataset from hypothesized values of the object's model parameters. The synthetic data are compared to the original data using the methods of probability theory. The comparison results in the probability that these particular parameter values could have given rise to the observed data. The probabilities are then used to update the model parameters. By repeating this sequence and iterating, one may automate the process of estimating model parameters from data.
The Bayesian methodology enables automation of parameter estimation for large datasets once synthetic data generation is accomplished. We describe the important concepts behind Bayesian modeling and illustrate the procedures with an application to Hubble Space Telescope images of planetary nebulae.
Previous | Session 153 | Next
Bulletin of the American Astronomical Society, 36 5
© 2004. The American Astronomical Society.