ABSTRACT
Gt =(a+ n)- 1 (aG0 + 2:.: 8w ). Interpreted in a different way, for a new 90 from G, P(90 E BIG) = G(B) so that, integrating over G, its marginal distribution is G0 . However, given 9*, P(90 E BJG)J9*) = G(B)J9* so that, again integrating over G, its conditional distribution is G~. That is, if 9'\', ... , 9~ and 96 are all drawn independently from G, then 96 is conditionally independent of 9* given G. However, we can "marginalize" over G (whence the 9/ are no longer independent since they shared a common random G) to obtain the conditional distribution of 9'/) given 9*. This distribution is Gb. We return to this point in Sec. 4.
