Application of Multivariate Statistical Analysis to Spectrum-Imaging Datasets: Benefits and Disadvantages

Masashi Watanabe^a, Kazuo Ishizuka^b

^aDepartment of Materials Science and Engineering, Lehigh University
^bHREM Research Inc.

Abstract: Multivariate statistical analysis (MSA) is one of essential approaches to effectively analyze large scale spectrum imaging (SI) datasets, which can be acquired in modern analytical electron microscopes (AEMs). In this article, first, principles and advantages of the MSA approaches based on the most popular principal component analysis (PCA) will be explained with several applications. Then, some issues/artifacts that might be introduced by applying the PCA method are also addressed. Finally, the recently developed LocalPCA method is also introduced，which has been developed by the authors to overcome some of artifacts introduced by the PCA.

Key words: spectrum imaging method, multivariate statistical analysis, principal component analysis, detectability limit