Parameter estimation of base-age invariant site index models: which data structure to use? - A discussion

We discuss here methods of comparing various approaches to site-dependent height-age model-parameter estimation. The discussion emphasis is on proper analysis of the varying parameter (VP) method versus several variants of the algebraic difference methods including one with the error in variable (EI...

Full description

Saved in:
Bibliographic Details
Published in:Forest science 2007-10, Vol.53 (5), p.552-555
Main Authors: Cieszewski, C.J, Strub, M
Format: Article
Language:English
Subjects:
accuracy
decision making
difference method
dummy variable
equations
estimation
forest growth
forest mensuration
forest yields
Forestry
Height
mathematical models
measurement
methodology
Methods
Parameter estimation
Pinus taeda
site index models
tree and stand measurements
varying parameter
Citations: Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We discuss here methods of comparing various approaches to site-dependent height-age model-parameter estimation. The discussion emphasis is on proper analysis of the varying parameter (VP) method versus several variants of the algebraic difference methods including one with the error in variable (EIV) theory adaptation. The study has demonstrated that, when properly considered, the VP method consistently outperforms all the other methods including the EIV method. The EIV method was tested on data with known "true model" and drastically simplified errors, and had a root mean square error (RMSE) more than 10 times worse than the VP method. All the other methods produced drastically and uniquely differing results with RMSEs varying from 26 times greater than the VP method to 300 times greater. The patterns of relatively close similarity between the EIV and the VP methods and the large differing among the other methods was similar for fittings to real data of second-rotation loblolly pine (Pinus taeda L.) from a designed experiment. The VP method is recommended for use over all the other methods as the most unbiased and as much simpler than the EIV method, which was second best.
ISSN:0015-749X
1938-3738
DOI:10.1093/forestscience/53.5.552