2024 Volume 80 Issue 20 Article ID: 24-20001
The availability of large, high-resolution data is continuously increasing in various areas of infrastructure planning. Model selection criteria used in statistics tend to support more complex models with more variables as the number of samples increases. However, this may not always be consistent with the direction of solution of the problems that our field expects in model estimation. In other words, for our specific problems, it is not clear how the complexity of the model should be captured according to the scale of the data, and what issues exist in such cases. This paper therefore discusses the following, using the estimation of location dependent fundamental diagrams as an example, which has been previously written by the author: the issues related to the data are summarised, the current state of the art of statistical model selection criteria is explained, the properties of the criteria are illustrated with examples, and future perspectives are presented.
