Noisy independent component analysis as a method of rotating the factor scores

  • Authors:

  • Steffen Unkel;Nickolay T. Trendafilov

  • Affiliations:

  • Department of Statistics, Faculty of Mathematics & Computing, The Open University, United Kingdom;Department of Statistics, Faculty of Mathematics & Computing, The Open University, United Kingdom

  • Venue:
  • ICA'07 Proceedings of the 7th international conference on Independent component analysis and signal separation
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

Noisy independent component analysis (ICA) is viewed as a method of factor rotation in exploratory factor analysis (EFA). Starting from an initial EFA solution, rather than rotating the loadings towards simplicity, the factors are rotated orthogonally towards independence. An application to Thurstone's box problem in psychometrics is presented using a new data matrix containing measurement error. Results show that the proposed rotational approach to noisy ICA recovers the components used to generate the mixtures quite accurately and also produces simple loadings.