Multivariate statistical analysis methods and their application in tobacco science

The paper was intended to open a multivariate statistical analysis (MSA) window where researchers, practitioners and managers capture some systemic pictures of MSA from a view of comprehensibility. Importance, essentiality and feasibility were analyzed of MSA application to tobacco science. Five MSA methods and their applications in applied botany including tobacco were introduced in a concise way, plus advances in the methods. These methods include (1) graphs and visualization of multivariate data; (2) principle component analysis for dimension reduction; (3) cluster analysis for classification and categorization; (4) discrimination analysis for identification and diagnosis; (5) canonical correlation analysis for interdependence and dependence between 2 sets of variables. Three additional traditional MSA methods of canonical variate analysis, multivariate variance analysis and correspondence analysis were listed by means of one sentence summary of their functions. New MSA methods were also mentioned of geo-statistics, support vector machine, projection pursuit, classification and regression trees, partial least square and structural equation modeling. Bootstrap resampling technique was strongly recommended to use together with MSA in order to enhance the reliability of results. ©, 2014, State Tobacco Monopoly Bureau and China Tobacco Society. All right reserved.