{"id":634,"date":"2017-09-28T06:50:45","date_gmt":"2017-09-28T06:50:45","guid":{"rendered":"http:\/\/jsr.isrt.ac.bd\/?post_type=article&p=634"},"modified":"2017-09-28T06:50:53","modified_gmt":"2017-09-28T06:50:53","slug":"predictive-distribution-heteroscedastic-multivariate-linear-models-elliptically-contoured-error-distributions","status":"publish","type":"article","link":"http:\/\/jsr.isrt.ac.bd\/article\/predictive-distribution-heteroscedastic-multivariate-linear-models-elliptically-contoured-error-distributions\/","title":{"rendered":"The predictive distribution for the heteroscedastic multivariate linear models with elliptically contoured error distributions"},"content":{"rendered":"
This paper considers the heteroscedastic multivariate linear model with errors
\nfollowing elliptically contoured distributions. The marginal likelihood function of
\nthe unknown covariance parameters and the predictive distribution of future responses
\nhave been derived. The predictive distribution obtained is a product of m
\nmultivariate Student\u2019s t distributions. It is interesting to note that when the models
\nare assumed to have elliptically contoured distributions the marginal likelihood
\nfunction of the parameters as well as the predictive distribution are identical to
\nthose obtained under independently distributed normal errors or dependent but
\nuncorrelated Student\u2019s errors. Therefore, the distribution of future responses
\nis unaffected by a change in the error distribution from the multivariate normal
\nand multivariate t distributions to elliptically contoured distributions. This gives
\ninference robustness with respect to departure from the reference case of independent
\nsampling from the multivariate normal or dependent but uncorrelated
\nsampling from multivariate t distributions to elliptically contoured distributions.<\/p>\n