search

statnet / ergm

Fit, Simulate and Diagnose Exponential-Family Models for Networks
Other
98 stars 37 forks source link

In control.ergm(SA.*) parameters, set defaults, including functional ones #511

Closed krivit closed 1 year ago

krivit commented 1 year ago

Currently, many of the control.ergm()'s SA.* arguments are set to NULL and substituted in ergm.stocapprox() .

Instead, they should be set to the defaults there. Where they are specific to the model or similar, they should be functions along the lines of init.MPLE.samplesize=.

krivit commented 1 year ago

Also, in the controls set by ergm.stocapprox(), use more descriptive names. For example, there is no reason to replace SA.nsubphases with nsub as it currently does.

AryaKarami commented 1 year ago

In the stoch.approx(), Is the argument "s.obs" related to the missing data? Why does it use !is.null(s.obs) ?

krivit commented 1 year ago

s.obs is an ergm_state object that is only non-NULL if there is a missing data process. It's passed because missing data support for SA is on the roadmap.