The real statement is that when the feature size is below the mean free path, inelastic scattering events become more rare, so carriers behave more and more ballistic. Scattering doesn't disappear completely, it's just a question of probability. As such the only way to estimate the mean free path is to compute the current with and without scattering, for different sizes of the system, and also for many configurations in case the scattering is random in nature, like with alloy scattering or dopant scattering. The link in Dipankar's answer is an excellent example of this, although it only covers dopant scattering - for a real estimate of the mean free path you would also need to consider inelastic scattering which is typically by phonons. This is possible in ATK 2015, but quite time-consuming, esp. for large systems (and remember, you need to make the system larger and larger to see when the effect becomes important).

Therefore, In most practical situations the mean free path is a lot easier to estimate in experiments, and then you can just choose to say that you perform simulations in a ballistic approximation in systems that are below that size. However, very interestingly, in some cases phonon scattering can be important no matter the size of the system. An example of such a calculation (for a tunnel FET) can be seen in

http://quantumwise.com/publications/tutorials/item/837,