[Bugs-r-devel] [Fwd: R2OpenBUGS mis-reports pD and DIC]

[Rodney S. <rsp...@mc...>]

Subject:R2OpenBUGS mis-reports pD and DIC

From:
Timothy Handley <Tim...@NP...>
Date:
Fri, 3 Feb 2012 10:16:46 -0600


I ran a model using R2OpenBUGS. The latter bit of output was:

   b.pcov.sd.u[1]           1.3  0.1    1.2    1.3    1.3    1.4    1.5
   1  1600
   c.Nzero.pre            142.3  6.3  129.0  138.0  142.0  147.0  154.0
   1  1800
   deviance              1175.0 24.6 1129.0 1158.0 1174.0 1191.0 1225.0
   1  1800

   For each parameter, n.eff is a crude measure of effective sample size,
   and Rhat is the potential scale reduction factor (at convergence,
   Rhat=1).

   DIC info (using the rule, pD = Dbar-Dhat)
   pD = 1400.0 and DIC = 224.9
   DIC is an estimate of expected predictive error (lower deviance is
   better).

The pD value was shocking. This model has several random effects, but even
so, I expected no more than 200 effective parameters. On further thought, I
was concerned to see that mean deviance (1175) plus pD (1400) was not equal
to DIC (224.9).

So I reran the model with debug=TRUE. Then, from within the OpenBUGS GUI, I
opened the DIC tool, and asked for stats on the node corresponding to the
data. OpenBUGS reported the following:


Deviance information
         Dbar      Dhat     DIC       pD
ones     1175.0    950.1    1400.0    224.9

These numbers are in the range of believability. They also have the
expected relationship, with DIC=Dbar+pD.

So it seems that R2OpenBUGS reverses the values for pD and DIC. It prints
the value for DIC in the space for pD, and prints the value for pD in the
space for DIC.

Tim Handley
Research Assistant
Mediterranean Network, NPS
805-658-5759 (CHIS)
805-370-2396 (SAMO)

-- 
Rodney Sparapani, PhD  Center for Patient Care and Outcomes Research
Sr. Biostatistician               http://www.mcw.edu/pcor
4 wheels good, 2 wheels better!   Medical College of Wisconsin (MCW)
WWLD?:  What Would Lombardi Do?   Milwaukee, WI, USA