search

brianstock / MixSIAR

A framework for Bayesian mixing models in R:
http://brianstock.github.io/MixSIAR/
94 stars 75 forks source link

Model never fully passes the diagnostic output #114

Open RCBlackburn opened 7 years ago

RCBlackburn commented 7 years ago

Hello,

I am attempting to run a model for bison diet through time using two isotopes (15N and 13C). I have set up concentration dependency model time as a continuous factor. I have ran the model at least 10 times on the extreme setting and experimented with different burn in values. The best diagnostic output I have gotten is below:

Generally the Gelman diagnostic should be < 1.05 Out of 363 variables: 4 > 1.01 0 > 1.05
0 > 1.1

The Geweke diagnostic is a standard z-score, so we'd expect 5% to be outside +/-1.96 Number of variables outside +/-1.96 in each chain (out of 363):

        Chain 1 Chain 2 Chain 3

Geweke 1 0 31

The diagnostic PDF shows the model visually converging.

Every model I have ran produces very similar results in terms of the proportions of plant groups found in bison diet through time.

Because I keep getting similar results and visual convergence but different diagnostic outputs can I rely on this model? If not, are there any suggestions to make this model better?

Let me know if any more information is needed.

Thanks! Ryan Blackburn

raphael-lavoie commented 6 years ago

Is your mixture data within the source polygon?