ergm
ergm copied to clipboard
statnet
Fit, Simulate and Diagnose Exponential-Family Models for Networks
A number of proposals have accumulated, particularly in the C code, many of which are one or more of the following: * is inaccessible: there is no `Init*ErgmProposal.*()` function that...
- [ ] Allow `map` and `gradient` arguments to be specified as `purrr`-style formulas. - [x] Allow the user to flag the `gradient` as "linear", in which case its (constant)...
I am not a CD expert, but there appear to be at least two reasons for this: - unlike MCMC and SAN, CD may call the U_FN with a dyad...
I recently encountered an issue with the gw*degree terms. Pleas see the simple reproducible example below: library(ergm) data(sampson) ergm(samplike ~ edges + gwodegree(fixed = FALSE)) The output is: Error: In...
E.g., ``` r library(ergm) library(coda) mk.s
I was not able to figure out the meaning of the `levels` argument to many of the terms from the existing documentation. Left them as a TODO for now. Alternatively,...
Most convergence detection algorithms we use (Hotelling, Hummel, confidence) rely on simulated values of estimating functions equalling on average the observed. However, with a box constraint, such as GWESP or...
Right now, curved terms can define box constraints on their own parameters, but it also makes sense to allow the user to do so. E.g., `control.ergm()` parameters `lower = -Inf`...
Current `output=` options either return a `stats` `mcmc.list` (or matrix, if `simplify=TRUE`) or an object or list (possibly nested) of objects of specified type (`network`, edge list, `ergm_state`) of length...
At the moment, `ergm_MCMC_sample()` takes a list of `ergm_state`s which are then each evolved separately; but that can't be used to resume the simulation from `simulate()`. We probably need an...