Oxygen abundance methods in the SDSS: view from modern statistics

Our purpose is to find which is the most reliable one among various oxygen abundance determination methods. We will test the validity of several different oxygen abundance determination methods using methods of modern statistics. These methods include Bayesian analysis and information scoring. We will analyze a sample of $\sim$6000 $\hii$ galaxies from the Sloan Digital Sky Survey (SDSS) spectroscopic observations data release four. All methods that we used drew the same conclusion that the $T_e$ method is a more reliable oxygen abundance determination methods than the Bayesian metallcity method under the existing telescope ability. The ratios of the likelihoods between the different kinds of methods tell us that the $T_e$, $P$, and $O3N2$ methods are consistent with each other because the $P$ and $O3N2$ method are calibrated by $T_e$-method. The Bayesian and $R_{23}$ method are consistent with each other because both are calibrated by a galaxy model. In either case, the $N2$ method is an {\it unreliable} method.