An errors-in-variable model with correlated errors
2014
Nigh, Gordon Donald
"Engelmann spruce (Picea engelmannii Parry ex Engelm.) is a high-elevation species found in northwestern North America. Its importance is increasing as the interest in these high-elevation sites grows. Consequently, growth intercept models that predict site index for this species need to be developed. Previous growth intercept models were fit with nonlinear least squares regression. With this technique, the assumption that the "x" variable (i.e., height) in the fitting is known without error is violated, resulting in a bias. The aim of this study was to remove this source of bias with the errors-invariable method of moments fitting technique. This involved developing errors-in-variable method of moments estimators for growth intercept models and fitting these models to stem analysis data. Nonlinear least squares regression and the method of moments estimators were then compared to evaluate the significance of the fitting technique. The method of moments models and the regression resulted in almost the same predictions except at ages less than approximately breast height age 15, where the method of moments estimators gave lower site index predictions. The errors-in-variable method of moments should be used to fit growth intercept models because it eliminates a source of bias. Some areas for further research in this technique still exist."
显示更多 [+] 显示较少 [-]