Monday, November 21, 2011 - 11:00am
"Hierarchical Bayesian Fitting of Dust SEDs"
Abstract. I will provide an overview of Bayesian inference and hierarchical modeling, in the context of fitting dust spectral energy distributions (SEDs). I will discuss why commonly employed frequentist methods, such as Chi^2 fits, may provide erroneous parameter estimates. Frequentist methods may even produce strong but artificial correlations - these appear to be similar to observed trends yielding a negative correlation between the dust temperature and spectral index. Our recently developed hierarchical Bayesian SED fitting method rigorously accounts for statistical uncertainties, and therefore returns accurate parameter estimates without any artificial correlations. We have applied the hierarchical Bayesian fitting method to Herschel observations of a nearby starless core, and find a slight positive correlation between the dust spectral index and temperature. This contrasts the Chi^2 result showing an anti-correlation between those parameters, and is suggestive of grain growth in the dense interstellar medium.