Log-evidence of the four models for noisy data from two variables.

The log-evidence computed by nested sampling for each model using 25 noisy data points from two repressilator variables. For these example models, it was found that the use of data from two variables gives more valuable information than an increase in the quantity of data from one variable. Using the interpretation given by Kass and Raftery [54] the data provide decisively strong evidence for the repressilator.