Analysis of air pollution plant exposure data: The soft independent modelling of class analogy (SIMCA) and partial least squares modelling with latent variable (PLS) approaches.

Air pollution plant exposure experiments sometimes involve registration of several plant variables (dependent variables) as a function of several treatment variables (independent variables). Both independent and dependent variables might be correlated, and the number of variables might exceed the number of observations. Analyzing one variable at a time gives a risk of spurious results. The multivariate statistical methods, soft independent modelling of class analogy (SIMCA) and partial least squares modelling with latent variables (PLS) allow all variables to be analyzed simultaneously. These methods are presented and applied to data from open-top chamber experiments with O3, SO2, and NO2 fumigation of three varieties of Lolium multiflorum Lam. The results demonstrate the dependence of the plant response to plant variety, plant age, climate, and pollutant dosages.