Comments on three-way analyses used for batch process data

Recently, several papers have appeared concerning the use of three-way models for batch process data. In these papers a number of points are raised. This paper discusses some of these points and illustrates some pitfalls. More specifically, some theoretical aspects of using different three-way models for batch process data and some practical consequences are discussed. The topics of cross-validation and data preprocessing are also discussed. These issues will be discussed using small simulated examples and theoretical arguments.