8  Final remarks

Linear mixed models is a vast subject with many application areas and specific uses cases. The number of decision to make in terms of how to model the dependencies is large as well. These decisions can have large impact on the inference made based on these models.

This implies that these models needs to be applied consciously, that the modelling decisions need to be made on the basis of the data structure and the understanding of the implications the modelling choices not on which model delivers most attractive results.

Inference is not always straightforward. Technically (How to derive confidence intervals, p-values) and from an interpretation point of view. Mixed models thrive on large data sets. Be careful with strict inference on data sets with very few groups in a random effect.