Multivariate statistical methods in battery research

Abstract The use of multivariate statistical methods in battery research is developed with examples drawn from the literature and unpublished work by the authors. The techniques discussed may be described in general as, data reduction, cluster analysis and regression methods for prediction. Individually or collectively these represent the three main areas of interest to battery researchers. Data reduction permits the visualization of the relationship between samples which are characterized by multiple measured variables. Cluster analysis extends this process to examine any natural groupings existing in the samples, based on the variables measured, and multivariate prediction is a calibration technique permitting the modelling of complex non-linear systems.