tag:blogger.com,1999:blog-3240271627873788873.post9043504311192260514..comments2024-03-26T06:46:11.752-04:00Comments on Doing Bayesian Data Analysis: Autocorrelation in Bayesian ANOVA: Fixing the book's overkillJohn K. Kruschkehttp://www.blogger.com/profile/17323153789716653784noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-3240271627873788873.post-72207535500508921322011-07-26T10:54:52.696-04:002011-07-26T10:54:52.696-04:00Jay: I probably misunderstand your comment, but I ...Jay: I probably misunderstand your comment, but I wanted to be sure that the method used in the programs is clear. The underlying MCMC chain is completely <b>un</b>affected by the reparameterization, and the "a" parameters in the MCMC chain remain badly autocorrelated. It's just that we <i>assess</i> autocorrelation by considering the reparameterized "b" values, not the "a" values. <br /><br />With respect to contrasts, again those are computed <i>after</i> the MCMC chain is generated, and the contrasts have no influence on the MCMC chain. In the programs many different contrasts are examined, without concern for their orthogonality. In principle, we should check that the chains for the contrasts are not badly autocorrelated, but in practice they are not, so it's not a problem.John K. Kruschkehttps://www.blogger.com/profile/17323153789716653784noreply@blogger.comtag:blogger.com,1999:blog-3240271627873788873.post-19553577462841621922011-07-25T23:06:51.299-04:002011-07-25T23:06:51.299-04:00Nice. MCMC usually "likes" deviation con...Nice. MCMC usually "likes" deviation contrasts better than dummy coding because the X matrix is orthogonal. Similarly, if you want something like polynomial contrasts it would make a great deal of sense to set them up as orthogonal contrasts.Jay Verkuilenhttps://www.blogger.com/profile/07461798676830653869noreply@blogger.com