Research Seminar
Orthogonal grey simultaneous component analysis to distinguish common and distinctive information in coupled dataMartijn SchoutedenKU Leuven | |
| Abstract: | Often, data are collected consisting of different blocks that all contain information about the same entities (e.g., items, persons, situations). A major challenge is to reveal the mechanisms underlying each of those coupled data blocks and to disentangle in this regard common mechanisms underlying all data blocks and distinctive mechanisms underlying a single data block or a few blocks only. An interesting class of methods for such an approach is the family of simultaneous component methods. These methods yield components that maximally account for the variance in all data blocks. Unfortunately, however, they usually contain a mix of common and distinctive information. Recently, DISCO-SCA has been proposed as a method to tackle this problem by orthogonally rotating a simultaneous component solution towards a target matrix with a clear common and distinctive structure (Schouteden et al., 2010). Yet, in quite a few cases, to improve the interpretability of the components, it may be needed to impose the target structure on the component solution to a stronger degree, while maintaining the orthogonality restriction. For this purpose, we developed a novel method, starting from the framework of grey component analysis as proposed by Westerhuis et al. (2007). After describing this method, we will illustrate it with data stemming from cross-cultural psychology and systems biology. |
| Date: | Tue Mar 16, 12:15 pm - 1:15 pm |
| Place: | room 02.51 (Department of Psychology, Tiensestraat 102, 3000 Leuven) |