Munich Personal RePEc Archive

Estimating regressions and seemingly unrelated regressions with error component disturbances

Paolo, Foschi (2005): Estimating regressions and seemingly unrelated regressions with error component disturbances.

[img]
Preview
PDF
MPRA_paper_1424.pdf

Download (141kB) | Preview

Abstract

The estimation of regressions models with two-way error component disurbances, is considered for the case where both the random effects are non-spherically distributed. The usual approach that first transforms the effects into uncorrelated ones and then applies within and between transformations, cannot be conveniently applied. Here, it is proposed to revert this scheme by firstly applying the within and between transformations. This results in simple General Linear Model which can be partitioned into three smaller GLMs. Then, by exploiting the structure of the models and using the Generalized QR decomposition as a tool, a computationally efficient and numerically reliable method for estimating the regression parameters is derived. This estimation method is generalized to the case of a system of seemingly unrelated regressions.

UB_LMU-Logo
MPRA is a RePEc service hosted by
the Munich University Library in Germany.