A multivariate statistical approach integrating the absolute principal components score (APCS) and multivariate linear regression (APCS-MLR), along with structural equation modeling (SEM), was used to model the influence of water chemistry variables on chlorophyll a (Chl a) in Lake Qilu, a severely polluted lake in southwestern China. Water quality was surveyed monthly from 2000 to 2005. APCS-MLR was used to identify key water chemistry variables, mine data for SEM, and predict Chl a. Seven principal components (PCs) were determined as eigenvalues > 1, which explained 68.67% of the original variance. Four PCs were selected to predict Chl a using APCS-MLR. The results showed a good fit between the observed data and modeled values. with R(2) = 0.80. For SEM, Chl a and eight variables were used: NH(4)-N (ammonia-nitrogen), total phosphorus (TP), Secchi disc depth (SD), cyanide (CN), arsenic (As), cadmium (Cd), fluoride (F), and temperature (T). A conceptual model was established to describe the relationships among the water chemistry variables and Chl a. Four latent variables were also introduced: physical factors, nutrients, toxic substances, and phytoplankton. In general, the SEM demonstrated good agreement between the sample covariance matrix of observed variables and the model-implied covariance matrix. Among the water chemistry factors, T and TP had the greatest positive influence on Chl a, whereas SD had the largest negative influence. These results will help researchers and decision-makers to better understand the influence of water chemistry on phytoplankton and to manage eutrophication adaptively in Lake Qilu. (C) 2009 Elsevier B.V. All rights reserved.
Guo G, Jiang T, Wang Y, Gao W. Finding Multiple Object Instances with Occlusion, in 20th International Conference on Pattern Recognition, ICPR 2010, Istanbul, Turkey, 23-26 August 2010. IEEE Computer Society; 2010:3878–3881. 访问链接
A general model is proposed for flexibly estimating the density of a continuous response variable conditional on a possibly high-dimensional set of covariates. The model is a finite mixture of asymmetric student t densities with covariate-dependent mixture weights. The four parameters of the components, the mean, degrees of freedom, scale and skewness, are all modeled as functions of the covariates. Inference is Bayesian and the computation is carried out using Markov chain Monte Carlo simulation. To enable model parsimony, a variable selection prior is used in each set of covariates and among the covariates in the mixing weights. The model is used to analyze the distribution of daily stock market returns, and shown to more accurately forecast the distribution of returns than other widely used models for financial data.