Let \(n\) and \(k\) be the number of data points and the number of parameters in a model respectively (assume \(n\gg k\)). Then BIC is defined as:

$$BIC = log(n) k - 2 log(\hat{L}), $$

where \(\hat{L}\) denotes the maximised value of the likelihood function. The model with a minimum value of BIC is preferred over the other models. It is worth noting that in order to evaluate the quality of a model, further statistical tests are needed, e.g. a hypothesis test.