ABSTRACT
The development of a NIR method, where the use of chemometric approaches is routinely employed, is often considered to be quite unique relative to other traditional analytical methods. This is especially challenging since a primary benefit of NIR is its application on intact samples with no sample preparation, resulting in spectra being complicated by unuseful interference. In this chapter, we will discuss the use of an analytical quality by design framework to support model development and then describe the tools used to evaluate the model performance during use. We will close out the chapter with case studies showing how various types of common variability can affect model performance and how to detect it.
