Given these difficulties the Bayesian perspective seems attractive, but now one is faced with quantifying the prior odds. In fact, there is now sufficient background knowledge to allow this to be done rather objectively [Wachholder et al., 2004; Wakefield, 2007], and it can be argued that the major obstacle to widespread use of this paradigm is convention. Nevertheless, these approaches can only be validated through experience, and as a rule of thumb, naïve Bayesian analysis should result in similar conclusions to a frequentist analysis, if both are well calibrated.
