|
Activity Number:
|
176
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 13, 2002 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Business & Economics Statistics Section*
|
| Abstract - #301284 |
|
Title:
|
Comparative Analysis of the Reduced-rank Regression: A Case Study Using Soviet Union Expenditures Data
|
|
Author(s):
|
R. Hanumara*+ and Andrada Toma and John Burkett
|
|
Affiliation(s):
|
University of Rhode Island and University of Rhode Island and University of Rhode Island
|
|
Address:
|
Tyler Hall, Kingston, Rhode Island, 02881, USA
|
|
Keywords:
|
Reduced-rank regression; ; Neural networks;
|
|
Abstract:
|
Usually, multivariate linear regression studies do not consider assumptions on the rank of the regression coefficient matrix. However, in some applications the coefficients matrix is not of full rank, and specific estimation methods are proposed. One such application of reduced-rank regression model uses the data on expenditures from Soviet Union State budget. This research presents results in applying the reduced-rank regression model on a power-transformed data set and compares them with the previous results obtained from a logarithmic transformation. Also, Bayesian learning for neural networks models implemented by hybrid Monte Carlo and Gibbs sampling algorithms is applied to the data set. The comparison of full rank, reduced-rank and neural network models is made in terms of goodness of fit and predictive power.
|
