Status of HPDinterval
My intention is to modify the scale on which the estimates of the variance components are estimated and then allow a VarCorr method for the mcmcsamp object. The VarCorr method will allow specification of the scale. The point of the new scale is that it allows the variance components to be zero but also to switch from zero to nonzero during the MCMC process. It would be possible to specify an HPD interval on the logarithm scale but, natually, there would need to be some provision for the zeros in the chain. Regrettably it will be at least a week or two before I can resume coding and checking this representation and sampling method. Another project will dominate this week, which also is the last teaching week of our semester. On Mon, May 5, 2008 at 5:42 AM, John Maindonald
<John.Maindonald at anu.edu.au> wrote:
Dear Douglas - I'm wondering whether the output from HPDinterval() in the new version of lme4 will (if you have made this decision) in due course appear with the same format as in the present coda version. CIs for variances will continue to be presented on a logarithmic scale? Also, will the coda functions in due course be able to accept the new merMCMC objects? I am putting together a note at present, re code that appears in our text, to which this is relevant. Thanks John. John Maindonald email: john.maindonald at anu.edu.au phone : +61 2 (6125)3473 fax : +61 2(6125)5549 Centre for Mathematics & Its Applications, Room 1194, John Dedman Mathematical Sciences Building (Building 27) Australian National University, Canberra ACT 0200.