Hierarchical Segmented Regression Models with Application to Wood Density Analysis
Wood density is an important characteristic of wood and one which plays a major role in determining the strength of wood products. One quantity useful in calculating wood density is called area-increment, which is the cross-sectional area of a specific ring at a particular height. Area increment is derived from radial measures taken from X-ray scan data that extend from the pith to the inside and outside of a particular ring.
The focus of this study is the modeling of area increment as a function of a scaled measurement of the tree-height at which area-increment was determined from samples taken at various heights of 60 lodgepole pine trees at BC. Lodgepole pine is an important commercial species that is highly responsive to intensive management practices and is grown for a wide variety of wood products.
The relationship between area increment and scaled tree height is modeled as a hierarchical segmented regression model. Slopes of the segments vary over trees in this mixed-effect modeling framework; it is also of interest to determine whether any covariates are explanatory for variation observed. Maximum likelihood estimation is performed for inference concerning the model parameters and the model is assessed using a variety of graphical techniques.
This type of interdisciplinary work is a hallmark of our program in Applied Statistics at Simon Fraser University. For more information, please contact Li Xing (firstname.lastname@example.org) or her supervisor Charmaine Dean (email@example.com), Department of Statistics and Actuarial Science.