Transcript Slide 1
Multivariate analysis of turfgrass cell wall components and relationship with black cutworm larval develpment S.C. Hong and R.C. Williamson The Dept. Entomology Univ. Wisconsin-Madison Objectives To examine multi-dimensional relationship between cell wall components of turfgrasses and black cutworm larval development To compare multivariate analyses Turfgrass species and cultivars Creeping bentgrass (cbe) Reveillie (tbr): The hybrid of Kentucky bluegrass and Texas bluegrass Kentucky bluegrass – – – – – Challenger (chl) Julia (jla) Midnight (mid) Monopoly (mono) South Dakota (sd) Young vs. Old based on planting dates – Young: less than 60 d – Old: greater than 365 d Cell wall components Dry matter Acid detergent fiber (adf) Neutral detergent fiber (ndf) Lignin Ash Total nitrogen Leaf toughness Larval weight (lw) from non-choice feeding assay Multivariate analysis Fisher’s discriminant analysis Cluster analysis – Hierarchical clustering method – Kruskal’s non-metric multidimensional scaling Factor analysis Fisher’s discriminant analysis Objective: – To describe graphically the different features of observations from populations – To classify turfgrasses into groups based on collected variables. Cluster analysis Exploratory data analysis tool – Sorting different objects into groups in a way that the degree of association between two objects is maximal. Kruskal’s non-metric multidimensional scaling (MDS) Hierarchical clustering method (HCM) Kruskal’s non-metric multidimensional scaling Hierarchical clustering method Factor analysis Objective: – To discover if the observed variables can be explained in terms of a much smaller number of variables called ‘factors’ – To fix collinearity Regression analysis using the result of factor analysis Factor analysis Rotated estimated factor loadings and communality by principal component (PC) Factor1 Factor2 Communalities Dry matter ADF NDF Lignin -0.087 0.824 0.880 0.802 0.918 0.430 0.348 -0.092 0.85 0.86 0.89 0.65 Ash Total nitrogen Leaf toughness -0.275 -0.733 0.653 -0.648 -0.059 0.566 0.5 0.54 0.75 Factor analysis ntmlw: log-transformed larval weight ntmfa1: new data calculated from the first factor loadings and cell wall data ntmfa2: new data calculated from the second factor loadings and cell wall data Factor analysis Source Coefficient T value p-value Linear model 1 Factor 1 Factor 2 -0.15342 -0.06544 -5.213 -1.000 <0.0001 0.324 Linear model 2 Factor 1 -0.16917 -6.807 <0.0001 Fitted model Log(LW) = 2.44 – 0.17 fa1 Adjusted R-square 0.5251 Summary Discriminant and Cluster analysis – Useful to describe graphically the features of cell wall data – Discriminant analysis and two cluster analyses show similar grouping pattern. Group one: intratypes Group two: Young vs. old Summary Factor analysis – Remedy for collinearity among some variables (ADF, NDF, lignin, and leaf toughness) – Fitted model with larval weight has a negative coefficient for factor 1 (fa1). Acknowledgements United States Golf Association Amber Klawitter Dow AgroSciences