ARS | Publication request: MULTIVARIATE 3-WAY DATA ANALYSIS TO IDENTIFY SPECIFIC PLANTS IN THE DIETS OF FREE-RANGING HERBIVORES

Submitted to: Pittsburg Conference on Analytical Chemistry and Applied Spectroscopy
Publication Type: Abstract Only
Publication Acceptance Date: February 2, 2005
Publication Date: February 26, 2005
Citation: Sisk, C., Pezzotti, C., Obeidat, S., Rayson, G.D., Fredrickson, E., Estell, R.E., Anderson, D.M. 2005. Multivariate 3-way data analysis to identify specific plants in the diets of free-ranging herbivores [abstract]. Pittsburg Conference on Analytical Chemistry and Applied Spectroscopy, February 26-March 4, 2005, Orlanda, Florida. p. 70.

Technical Abstract: The issue of identifying specific plant species in the diets of free ranging herbivores in the local and global community is of much interest in the context of three-way analysis. The identification process was analyzed using chemometrics methods applied to the excitation emission fluorescence of a number of plant species. We were able to develop an integrated approach to perform tasks such as multidimensional analysis, process modeling and statistical process decomposition. A description is given to the advantages of combining several chemometrics tools (Parallel Factor Analysis (PARAFAC), multiway principal component analysis, PARAFAC II, and GRAM) and the principal benefits associated with each strategy for the elucidation of diet composition. Also a general overview of the principal achievements and limitations of the techniques used within the presented methodology is depicted. It is illustrated how three-way principal components analysis as the appropriate generalization of conventional principal component analysis may serve as a powerful method for classification of specific plant species in diets of free ranging herbivores using the excitation-emission matrices from fluorescence spectroscopy from different species. The factors found appear to correspond to the causal influences manipulated in the experiment, revealing their patterns of influence in all three ways of the data. Several generalizations of the parallel factor analysis model are currently under development, including ones that combine parallel factors with Tucker-like factor ‘interactions’. In the research, necessary and sufficient conditions for global and local solutions to plant identification are being derived. The results of these investigations will be presented and the implications of the application of these data analysis tools for the identification of specific noxious weeds within the diets of free-ranging cattle will be discussed.