Models comparisons

[choisy]

It would be very nice and useful to have a function that compares a set of models and looks for the best one in terms of bias and variance. The general gold standard is cross-validation. In case of parametric models, we can save time by using approximation metrics of cross-validation such as AIC, BIC, etc... And in case of parametric models that are nested (i.e. one being a particular simpler case of the other), comparison can be done statistically with a likelihood ration test (LRT).