2006MNRAS.371.1693E

The next generation of weak gravitational lensing surveys are capable of generating good measurements of cosmological parameters, provided that, amongst other requirements, adequate redshift information is available for the background galaxies that are measured. It is frequently assumed that photometric redshift techniques provide the means to achieve this. Here, we compare Bayesian and frequentist approaches to photometric redshift estimation, particularly at faint magnitudes. We identify and discuss the biases that are inherent in the various methods, and describe an optimum Bayesian method for extracting redshift distributions from photometric data.