UNIGE document Scientific Article
previous document  unige:16438  next document
add to browser collection

Isotone additive latent variable models

Published in Statistics and computing. 2011, vol. 22, p. 647-659
Abstract For manifest variables with additive noise and for a given number of latent variables with an assumed distribution, we propose to nonparametrically estimate the association between latent and manifest variables. Our estimation is a two step procedure: first it employs standard factor analysis to estimate the latent variables as theoretical quantiles of the assumed distribution; second, it employs the additive models' backfitting procedure to estimate the monotone nonlinear associations between latent and manifest variables. The estimated fit may suggest a different latent distribution or point to nonlinear associations. We show on simulated data how, based on mean squared errors, the nonparametric estimation improves on factor analysis. We then employ the new estimator on real data to illustrate its use for exploratory data analysis.
Keywords Factor analysisPrincipal component analysisNonparametric regressionBartlett's factor scoresDimension reduction
Full text
Article (Submitted version) (657 Kb) - public document Free access
(ISO format)
SARDY, Sylvain, VICTORIA-FESER, Maria-Pia. Isotone additive latent variable models. In: Statistics and computing, 2011, vol. 22, p. 647-659. doi: 10.1007/s11222-011-9262-z https://archive-ouverte.unige.ch/unige:16438

919 hits



Deposited on : 2011-06-27

Export document
Format :
Citation style :