Harmonizing biomass tables by generalized least squares

作者：Tiberius Cunia, Russell D. Briggs · 发表于：Canadian Journal of Forest Research · 年份：1985 · DOI：10.1139/x85-054 · 被引用次数：7 · 研究领域：Forest ecology and management、Environmental Impact and Sustainability、Forest Management and Policy

To construct biomass tables for various tree components that are consistent with each other, one may use linear regression techniques with dummy variables. When the biomass of these components is measured on the same sample trees, one should also use the generalized rather than ordinary least squares method. A procedure is shown which allows the estimation of the covariance matrix of the sample biomass values and circumvents the problem of storing and inverting large covariance matrices. Applied to 20 sets of sample tree data, the generalized least squares regressions generated estimates which, on the average were slightly higher (about 1%) than the sample data. The confidence and prediction bands about the regression function were wider, sometimes considerably wider than those estimated by the ordinary weighted least squares.