Accession Number : AD0765624

Title :   Estimating Heteroscedastic Variances in Linear Models - A Simpler Approach.

Descriptive Note : Technical rept.,

Corporate Author : JOHNS HOPKINS UNIV BALTIMORE MD DEPT OF MATHEMATICAL SCIENCES

Personal Author(s) : Duncan,David B. ; Horn,Susan D. ; Horn,Roger A.

Report Date : JUL 1973

Pagination or Media Count : 23

Abstract : The authors describe an estimator of heteroscedastic variances in the Gauss-Markov linear model 7 = X beta + epsilon where E(epsilon) = O and Var (epsilon) = diag((Sigma sub 1, Sup 2),..., (Sigma sub n, sup 2)) with (Sigma sub i, sup 2) and beta unknown. It may be thought of as an approximation to the MINQUE method, but it results in both computational economy and decreased mean square error. Properties of this approximately unbiased estimator are stated and it is compared with other estimators. Extensions to more general models are discussed. (Author)

Descriptors :   (*ANALYSIS OF VARIANCE, MATHEMATICAL MODELS), APPROXIMATION(MATHEMATICS), REGRESSION ANALYSIS, MATRICES(MATHEMATICS), SAMPLING, MATHEMATICAL PREDICTION

Subject Categories : Statistics and Probability

Distribution Statement : APPROVED FOR PUBLIC RELEASE