Effective number of parameters about mixsiar HOT 4 CLOSED

Yes. If you're running the GUI, the raw rjags output exists in the "mixsiar" environment. You can print it to the screen with:

mixsiar$jags.1

To output the pD by itself with:

mixsiar$jags.1$BUGSoutput$pD

from mixsiar.

The mixsiar$jags.1 command still gives summary statistics, can I access raw data for all iterations somehow to combine sources and determine the statistics myself?

from mixsiar.

mixsiar$jags.1 is the rjags object, which contains the MCMC chains. To access the MCMC chains (p.global, p.fac1, p.fac2, depending on the random/fixed effects you included), use:

attach.jags(mixsiar$jags.1)

Now if you type dim(p.global) you should see that p.global is a matrix, where the number of columns is n.sources, and the number of rows is chain.length*n.chains. Full chain for proportion source 2 (overall/global population):

p.global[,2]

If you have random/fixed effects in your model, you'll see that p.fac1 and p.fac2 are 3-D arrays, where the dimensions are chain.length*n.chains x factor level x source.

Full chain for proportion source 1, level 1 of factor 1 (Region 1 Deer in wolves ex):

p.fac1[,1,1]

Full chain for proportion source 3, level 8 of factor 2 (Pack 8 Salmon in wolves ex):

p.fac2[,8,3]

To confirm you understand the indexing correctly, quickly plot a histogram of the chain and compare it to the output plot, e.g.

hist(p.fac2[,8,3])

should match the blue curve in "posterior_density_diet_p_Pack 8.pdf".

from mixsiar.

Thank you very much Brian that worked perfectly.

from mixsiar.