Studentization and prediction problems in multivariate multiple regression
Titel:
Studentization and prediction problems in multivariate multiple regression
Auteur:
Guttman, Irwin Hougaard, P.
Verschenen in:
Communications in statistics
Paginering:
Jaargang 14 (1985) nr. 6 pagina's 1251-1258
Jaar:
1985
Inhoud:
The problem addressed in this paper is the following. Based on data Y =XB + E, where E is matric-normal, that is E N N N X P[O; IN, C], we wish to predict Yf , where Yf is to be generated-independently of the data Y, and such t h a t , given B, Xf, C Y =X B +Ef, where Ef $NNf[O; InfNff, El. (The subscript “f” stands for “future”, future to Y.) X and X are design matrices f of order (N x k) and (Nf x k) , respectively, with B of order (k xp). We obtain the predictived is tribution of If using a frequentist approach and contrast this with the known r e s u l t s of the Bayesian pre-posterior approach, and discuss the basicdifferences between the two approaches.