6.14. Odds-ratio’s#

There are a multitude of statistical measures that are presented as odds-ratios (hereafter OR). They are often seen in terms of evaluating clinical studies. For instance, the odds-ratio of statin medication increasing life-span. We will be concerned with using the odds-ratio to distinguish between competing hypotheses.

Consider an odds ratio that contrast the probability of a sequence under an alternate hypothesis (\(p_{alt}\)) against the probability of the sequence under a null hypothesis (\(p_{null}\)). In this instance, an OR equal to 1 if \(p_{alt} = p_{null}\), i.e. the sequence is equally likely under both models. OR > 1 indicates the sequence is more likely under the alternate, while a value less than 1 implies the converse.

6.14.1. Log-odds – the logarithm variant#

A log-odds ratio is simply the log-transformation of the odds ratio [1]. For example, for OR=1, the log-odds \(\log_{10}(OR)=0\) [2].