PARAMETER ESTIMATION OF NONLINEAR SPLITPLOT DESIGN MODELS: A THEORETICAL FRAMEWORK

  • I. J. David Department of Statistics, Ahmadu Bello University, Zaria, Nigeria
  • O. E. Asiribo Department of Statistics, Ahmadu Bello University, Zaria, Nigeria
  • H. G. Dikko Department of Statistics, Ahmadu Bello University, Zaria, Nigeria
Keywords: Nonlinear Split-Plot Design Model, Gauss-Newton, Estimated Generalized Least Square, Restricted Maximum Likelihood Estimation

Abstract

Split-plot design models are special class of linear models with two error terms, that is the whole plot error and subplot error terms. These models can be remodeled as nonlinearly with variance components. This is a combination of nonlinear model for the mean part of the split-plot design model with additive error terms which describes the covariance configuration of the models. This research work presents estimated generalized least square method for estimating the parameters of the nonlinear split-plot design models. To achieve this, an iterative Gauss-Newton procedure with Taylor Series expansion was implemented. The unknown variance components of the model are estimated via residual maximum likelihood estimation method. The advantage of this technique is that it produces stable numerical values for the parameters mean and variances since it considers the covariance configuration of the model.

 

Downloads

Download data is not yet available.
Published
2019-05-11
Section
Articles