MCMC diagnostics section for dyadic independent models

[smjenness]

When a model containing only dyadic independent terms is fit (for example, ~edges + nodematch("Grade") for the add health datasets), no MCMC is run as expected. But it may be helpful for the novice user to have some explanation of this in the MCMC diagnostics tab with such models.

Another option is to always force MCMC even with dyadic independent models with the control.ergm setting force.main set to TRUE, which would always run MCMC. But I'm not sure if this is a good statistical practice.

[skyebend]

I ran into the same issue: It wasn't obvious why I see the "MCMC was not run or MCMC sample was not stored" mesesage on the MCMC Diagnostics window. (I'm assuming because this was a dyadic independent model, so it didn't need to run it, just did MLE, right?) I was thinking, "OK, where is the 'Run MCMC' button?

[martinamorris]

no MCMC for DI models, but that could be a bit clearer...

On Mon, 3 Nov 2014, Skye Bender-deMoll wrote:

I ran into the same issue: It wasn't obvious why I see the "MCMC was not run or MCMC sample was not stored" mesesage on the MCMC Diagnostics window. (I'm assuming because this was a dyadic independent model, so it didn't need to run it, just did MLE, right?) I was thinking, "OK, where is the 'Run MCMC' button?

— Reply to this email directly or view it on GitHub.[AH0J4jdyq_cX5L-YC6m4wJW1cMOQEhIDks5nJ8_ngaJpZM4C1-BZ.gif]

---

Professor of Sociology and Statistics Director, UWCFAR Sociobehavioral and Prevention Research Core Box 354322 University of Washington Seattle, WA 98195-4322

Office: (206) 685-3402 Dept Office: (206) 543-5882, 543-7237 Fax: (206) 685-7419

morrism@u.washington.edu http://faculty.washington.edu/morrism/

[ebey]

I changed the output message when MCMC doesn't run to be "MCMC does not run for dyadic independent models". Is more explanation necessary? Perhaps about what dyadic independence means?

[martinamorris]

if possible, we should use the short default output directly, with a help symbol that brings up this text:

"MCMC is only run when at least one of the terms in the model represents dyad dependence (e.g., degree terms, or triad related terms). For models with only dyadic independent terms, estimation relies on traditional maximum likelihood algorithms used for generalized linear models."

On Mon, 3 Nov 2014, ebey wrote:

I changed the output message when MCMC doesn't run to be "MCMC does not run for dyadic independent models". Is more explanation necessary? Perhaps about what dyadic independence means?

— Reply to this email directly or view it on GitHub.[AH0J4lWtk6FQUhWNsHDr0EmlbjkaZE1nks5nJ9NKgaJpZM4C1-BZ.gif]

---

Professor of Sociology and Statistics Director, UWCFAR Sociobehavioral and Prevention Research Core Box 354322 University of Washington Seattle, WA 98195-4322

Office: (206) 685-3402 Dept Office: (206) 543-5882, 543-7237 Fax: (206) 685-7419

morrism@u.washington.edu http://faculty.washington.edu/morrism/

[ebey]

Fixed in commit 5e8cc0a8