Accounting for Calibration Uncertainties in X-ray Analysis: Effective Areas in Spectral Fitting

Hyunsook Lee
Harvard-Smithsonian Center for Astrophysics

ABSTRACT:
Accounting for calibration uncertainties when using existing x-ray analysis packages requires painstaking extensive case specific simulations, whereas ignoring these uncertainties underestimates error bars. In this talk, we present two general statistical methods that incorporate calibration uncertainties. One is based on multiple imputation of which procedure can be applied with any fitting analysis tools by combining variances dues to statistical and systematical uncertainties. The other is a Bayesian approach that aggregates calibration uncertainties into a statistical model fit via Markov-chain Monte Carlo. In both cases, computational efficiency is improved by summarizing calibration uncertainty with a principle component analysis of simulated calibration files that span the systematic variations. These methods are implemented using recently codified Chandra effective area uncertainties and are verified using both simulated and actual Chandra ACIS-S data. The significance of this study is that the procedure of incorporating the effective area uncertainty is easily generalized for other types of calibration uncertainties.