Presentation 1-1, 2:50PM~3:20PM JST.
Speaker
Koji Tsukuda (Kyushu University)
Title
A study on estimation in the multivariate allometric regression model.
Abstract
The allometric regression model is a multivariate multiple regression model imposing that the difference between expectation vectors of two observations with different values of explanatory variables is parallel to the first principal eigenvector. In this presentation, estimation of the first principal eigenvector in the allometric regression model is discussed. Considering a class of estimators based on a weighted matrix of the regression sum of square matrix and the residual sum of squares matrix, we propose a new estimator that purposes to decrement an upper bound of the mean squared error of an estimator contained in this class. The proposed estimator and some conventional estimators are numerically compared.
This presentation is based on the preprint: Tsukuda and Matsuura (2024, arXiv:2402.11219).