Linear Tree Growth Models: a Limitation of Negative Estimation of Size

Abstract

The use of linear tree growth models, to describe the change in the size/response variate viz volume, biomass) with respect to the explanatory variate (height, diameter at breast height, age), is very common in forestry/agroforestry. However, these models may sometimes lead to misleading predictions viz 'negative estimation of size' exclusively for lower range of the explanatory variate. This study elaborates this issue and a feasible solution is proposed to overcome this problem. It is contrived that when the phenomenon is inherently undesirable, as the negative estimation of size in simple linear models for lower range of explanatory variable, to put implied lower bound on the raw values is most appropriate.