welcome: please sign in
location: CategoryBugComputational / PoissonSamplingDifferences

Brief Description

I have an annual demographic model of a breeding bird population. I have data on all demographic parameters (abundance, productivity, survival) except one, the number of breeding attempts (expected to be between 1 and 2). So I model annual values of this (rho) as drawn from a Poisson distribution with a given mean (nb_mean). I would expect the annual values of rho to vary on each sampled occasion, but in openbugs they do not, being fixed, usually at 1 (thus with an se equal zero). Winbugs seems to provide much more sensible samples, ie they vary on each sampling. Interestingly, if I have a (much) more complicated version of the model, openbugs will provide 'sensible' (ie se not equal 0) for some of the annual estimates of rho.


*Operating System: Windows XP x64


*OpenBUGS version: 3.1.0 rev482

Example

GUI instructions or script (preferred) to produce bug (Use attachment button to add)

modelCheck('model1.txt') modelData('data1.txt') modelCompile(1) modelInits('initial1.txt', 1) samplesSet(rho) modelUpdate(1000) samplesStats("rho")


Model, data, inits (.odc file preferred, Use attachment button to add)

Note: these files are cut down versions of the full model so parameters estimated in the full model (which spans 20 years) are represented here as constants close to their mean estimated value.

model1.txtinitial1.txtdata1.txt


Plan


CategoryBugComputational/PoissonSamplingDifferences (last edited 2010-07-05 16:11:58 by RobRobinson)