Journal of the Japan Statistical Society, Japanese Issue
Online ISSN : 2189-1478
Print ISSN : 0389-5602
ISSN-L : 0389-5602
ON A VARIABLE SELECTION PROCEDURE FOR A MODIFIED LIMITED INFORMATION MAXIMUM LIKELIHOOD METHOD
Haruo Onishi
Author information
JOURNAL FREE ACCESS

1985 Volume 15 Issue 1 Pages 35-44

Details
Abstract
K. Morimune proposed a modified limited information maximum likelihood (MF-LIML) which possesses exact moments and is asymptotically unbiased to O (T-1). The j-th best subset problem is formulated for the variable selection problem of MF-LIML. The solution to the first best subset problem (j=1) or the ultimately best subset among the J solutions obtained by solving the first to the J-th best subset problems (e. g., J=10) in one computer-run is proposed as a pragmatical solution to the variable selection problem for MF-LIML. The proposed variable selection procedure can make the laborious work of estimation by MF-LIML less time-consuming, labor-consuming and costly.
Content from these authors
© Japan Statistical Society
Previous article Next article
feedback
Top