Regression Smoothers for Estimating Parameters of Growth Analyses

This paper describes and evaluates two recent advances in regression smoothers in the context of plant growth analysis and compares them to natural B-splines and shows that these smoothers can capture even subtle changes in relative growth rate.

Abstract

The objective of regression smoothers is to obtain predicted values of a dependent variable and its first derivative from empirical data without having to assume any particular functional relationship between the dependent and independent variables. An early variant of this type of analysis, specifically natural B -splines, was first applied to growth analyses by Parsons and Hunt in 1981 ( Annals of Botany48 : 341–352, 1981). The object of this paper is to describe and evaluate two recent advances in this area (cubic spline smoothers and loess smoothers) in the context of plant growth analysis and compare them to natural B -splines. The accuracies of these methods are evaluated using simulated data of a type that normally causes difficulties with other methods. A bootstrap procedure is described that improves the estimate of the optimal smoother parameter. It is shown that these smoothers can capture even subtle changes in relative growth rate. The method is then applied to growth data of Holcus lanatus.