Decomposed Triangular Matrices on the Estimators of Unequal Observations of Seemingly Unrelated Regression Model
Loading...
Date
2018-03
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Journal of the Nigerian Association of Mathematical Physics
Abstract
The seemingly Unrelated Regression model is a generalisation of a linear regression model that is linked by contemporaneously correlated error terms. This work examined the effects of unequal sample observations on the partitioned triangular matrix in seemingly unrelated regression using Cholesky method of decomposition. The partitioned upper and lower non-singular triangular matrices was to establish contemporaneous relationship among equations, through their errors. A Monte Carlo experiment was performed on a three-equation model with sample sizes n = 20, 25, 40, 50, 60, 75, 80 and 100 for equal observations and T = 20, 40, 60 and 80, with extra observations E = 5, 10, 15 and 20 respectively for the unequal observations. The Root Mean Square Error (RMSE) was used to adjudge the performance of the estimators when unequal observations are considered. Unequal observations had little or no effect on both the upper and lower triangular matrices in this experiment. It was also observed that RMSE of SUR estimator was lower than that of OLS estimator for both equal and unequal number of observations, irrespective of the triangular matrix used. The upper triangular matrix had higher RMSE values than the lower triangular matrix for SUR and OLS estimators.
Description
Keywords
Extra Observations, Triangular matrices, Root Mean Square Error, Ordinary Least Squares, Seemingly Unrelated Regression