Binomial-reference ERGM with multiple parameters

statnet / ergm.count

Fit, Simulate and Diagnose Exponential-Family Models for Networks with Count Edges

Other

10 stars 1 forks source link

Binomial-reference ERGM with multiple parameters #3

Closed allenkei closed 3 years ago

allenkei commented 3 years ago

Hi,

Thanks for the great package again!

The paper, Exponential-Family Random Graph Models for Valued Networks, provides a close form Poisson-reference ERGM with network sum as statistics. Just wondering for Binomial-reference ERGM, when \theta is a vector of multiple parameters for multiple network statistics, does Y_{ij} still follow a Binomial(m, logit^{-1}(\theta))? What is the distribution in close form for a Binomial-reference ERGM with multiple parameters?

Thank you so much in advance! Allen

krivit commented 3 years ago

It will remain binomial for GLM-style effects: the ones that are dyad-independent terms for binary ERGMs.

allenkei commented 3 years ago

Dear Dr. Krivitsky,

Thank you very much and Merry Christmas!

Allen

allenkei commented 3 years ago

Dear Dr. Krivitsky,

Please forgive me to follow up with another question. I am writing to confirm if I have understand this correctly.

For a Binomial-reference valued ERGM with 2 network statistics,

that is each dyad in this network follows a Binomial distribution, with the same number of trials (m) and the probability of success in each trial is (\pi). Is this formulation correct for a Binomial-reference valued ERGM?

Thank you so much again and wish you a happy New Year!

Allen

CarterButts commented 3 years ago

@allenkei I would suggest directing these to the statnet help list, or contacting folks directly. Posting these questions here sends them to everyone on the development team as software issues, which is not ideal.