Abstract
In virtually all applications of factor models to aerosol mass apportionment, two important implicit assumptions are made: (a) that the source composition profile matrices remain constant over samples, and (b) that the (diagonal) error covariance matrices are the same from sample to sample. As observed by Gleser [Chemom. Intell. Lab. Syst., 10 (1991) 45] and Currie [Chemom. Intell. Lab. Syst., 10 (1991)], neither of these assumptions is generally true. In this paper, a factor analysis model is presented that accounts for variation over samples of both the source composition profiles and the error covariance matrices. Based on this model, point estimators and confidence set estimators are obtained for the vectors of source contributions using an empirical Bayes approach.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call