Bayesian and Non-Bayesian Tests of Independence in Seemingly Unrelated Regressions
推导了检验两组似不相关回归是否独立的贝叶斯与非贝叶斯统计量，包括似然比检验、沃尔德检验、拉格朗日乘子检验和贝叶斯检验，并通过抽样实验比较了四种检验的表现。
Bayesian and non-Bayesian statistics are derived for testing whether or not two blocks of seemingly unrelated regressions are independent. The non-Bayesian statistics are the likelihood ratio test (LRT), Wald's test (WT), and the Lagrange multiplier test (LMT). The authors interpret the LMT and WT as differences in the log of likelihoods conditioned on estimates of nuisance parameters. The Bayesian test is a highest posterior density region test that can also be derived as a Bayes factor. The four tests are compared in sampling experiments. Copyright 1988 by Economics Department of the University of Pennsylvania and the Osaka University Institute of Social and Economic Research Association.