The applications of projection pursuit (PP) to some real data sets are described. Some applications have been published in subject-matter scientific journals and have a straightforward interpretation within the related fields. Other data sets are well-known in the statistical literature. For example, kurtosis minimization sequentially recovers the cluster structure of Fisher Iris Data. The results obtained with PP are compared with those obtained with other dimension reduction methods, for example, principal component analysis and invariant coordinate selection. In all the addressed applications, PP is based on either skewness or kurtosis optimization. The related algorithms are implemented in the R packages Kurt, MaxSkew and MultiSkew.
Projection pursuit: An empirical tour
Cinzia FranceschiniSoftware
;Nicola Loperfido
Methodology
2022
Abstract
The applications of projection pursuit (PP) to some real data sets are described. Some applications have been published in subject-matter scientific journals and have a straightforward interpretation within the related fields. Other data sets are well-known in the statistical literature. For example, kurtosis minimization sequentially recovers the cluster structure of Fisher Iris Data. The results obtained with PP are compared with those obtained with other dimension reduction methods, for example, principal component analysis and invariant coordinate selection. In all the addressed applications, PP is based on either skewness or kurtosis optimization. The related algorithms are implemented in the R packages Kurt, MaxSkew and MultiSkew.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
