B0958
Title: Nonparametric estimation in the source apportionment problem
Authors: Jordan Bryan - Duke University (United States) [presenting]
Abstract: Motivated by inference questions arising in water quality assessment, a simple procedure is proposed for estimating quantities of dissolved organic nitrogen (DON) using fluorescence spectroscopy data. In particular, the source apportionment problem is studied, in which the composition of DON in a river is assumed to be determined by the land-use sources in the river's vicinity. The estimator utilizes excitation-emission matrices (EEMs) measured at several land-use sources to estimate the contribution of each source to the river's total DON profile. Although the estimator can be described succinctly by ordinary least squares (OLS) regression, it is demonstrated that it has connections to generalized least squares (GLS) and empirical Bayes methods. It is also shown that it performs favorably compared to other estimation strategies on a dataset of EEMs collected from the Neuse River system in North Carolina.
