Large-sample properties of unsupervised estimation of the linear discriminant using projection pursuit
Radojičić, U., Nordhausen, K., & Virta, J. (2021). Large-sample properties of unsupervised estimation of the linear discriminant using projection pursuit. Electronic Journal of Statistics, 15(2), 6677-6739. https://doi.org/10.1214/21-EJS1956
Published inElectronic Journal of Statistics
© Authors, 2021
We study the estimation of the linear discriminant with projection pursuit, a method that is unsupervised in the sense that it does not use the class labels in the estimation. Our viewpoint is asymptotic and, as our main contribution, we derive central limit theorems for estimators based on three different projection indices, skewness, kurtosis, and their convex combination. The results show that in each case the limiting covariance matrix is proportional to that of linear discriminant analysis (LDA), a supervised estimator of the discriminant. An extensive comparative study between the asymptotic variances reveals that projection pursuit gets arbitrarily close in efficiency to LDA when the distance between the groups is large enough and their proportions are reasonably balanced. Additionally, we show that consistent unsupervised estimation of the linear discriminant can be achieved also in high-dimensional regimes where the dimension grows at a suitable rate to the sample size, for example, pn=o(n1∕3) is sufficient under skewness-based projection pursuit. We conclude with a real data example and a simulation study investigating the validity of the obtained asymptotic formulas for finite samples. ...
PublisherInstitute of Mathematical Statistics
Publication in research information system
MetadataShow full item record
Additional information about fundingThe work of Joni Virta was supported by the Academy of Finland (Grant 335077).
Showing items with similar title or keywords.
Niku, Jenni; Brooks, Wesley; Herliansyah, Riki; Hui, Francis K. C.; Taskinen, Sara; Warton, David I. (Public Library of Science, 2019)Generalized linear latent variable models (GLLVM) are popular tools for modeling multivariate, correlated responses. Such data are often encountered, for instance, in ecological studies, where presence-absences, counts, ...
Hentunen, Henri (2019)Today’s fragmented and constantly changing business environment pushes companies more and more to forecast their future performance as accurately as possible. One approach for forecasting future can be done with estimation. ...
Fashina, Alfred (2021)Estimating the cost, effort, and size to complete a software project is one of the most difficult and confusing tasks confronted by software project managers. Though, an early estimate is very crucial when bidding for ...
Selenius, Elli (2020)Metal clusters are nanoparticles that have from two to thousands of metal atoms. The properties of metal clusters are extremely size-dependent, and adding or removing even one atom can make a difference. The optical ...
Duran, Derya; Jacknick, Christine M. (Elsevier, 2020)This paper explores teacher elicitation practices following a perceived absence of a response to an initial inquiry. Specifically, we focus on whole class post-task discussions where a teacher pursues responses in post-first ...