simulate_longitudinal_network: A function to simulate longitudinal directed networks using a...

In ctross/STRAND: An R package for simulating and analyzing social network data in R and Stan

A function to simulate longitudinal directed networks using a combined stochastic block and social relations model

Description

This is a function to simulate multilayer longitudinal network data with a stochastic block structure, sender-receiver effects, and dyadic reciprocity. This function is essentially the union of a social relations model and a stochastic block model.

Usage

simulate_longitudinal_network(
  N_id = 99,
  N_timesteps = 3,
  B = NULL,
  V = 3,
  groups = NULL,
  sr_mu = c(0, 0),
  sr_sigma,
  sr_Rho,
  dr_mu = 0,
  dr_sigma,
  dr_Rho,
  outcome_mode = "bernoulli",
  link_mode = "logit",
  individual_predictors = NULL,
  dyadic_predictors = NULL,
  individual_effects = NULL,
  dyadic_effects = NULL
)

Arguments

N_id	Number of individuals.
N_timesteps	Number of network layers, one per timestep.
B	List of List of matrices that hold intercept and offset terms. Log-odds. The first matrix should be 1 x 1 with the value being the intercept term. The first list is over layers/time-steps. The second is over block variables within layers.
V	Number of blocking variables in each layer.
groups	A list of dataframes of the block IDs of each individual for each variable in B.
sr_mu	A vector for sender and receivier random effects. In most cases, this should be a vector of 0s 2*N_timesteps long.
sr_sigma	A standard deviation vector for sender and receivier random effects. The first N_timesteps elements control node-level variation in out-degree, the second N_timesteps elements control in-degree.
sr_Rho	Correlation of sender-receiver effects (i.e., generalized reciprocity). Needs to be a valid correlation matrix, 2*N_timesteps by 2*N_timesteps.
dr_mu	A vector for dyadic random effects. In most cases, this should be a vector of 0s N_timesteps long.
dr_sigma	Standard deviation for dyadic random effects. This should be a vector N_timesteps long.
dr_Rho	Correlation of dyad effects (i.e., dyadic reciprocity). Needs to be a valid correlation matrix, 2*N_timesteps by 2*N_timesteps.
outcome_mode	Outcome mode: can be "bernoulli", "poisson", or "binomial".
link_mode	Link mode: can be "logit", "probit", or "log". For pois, you must use log.
individual_predictors	An N_timesteps list of N_id x N_individual_parameters array of covariates.
dyadic_predictors	An N_timesteps list of N_id x N_id x N_dyadic_parameters array of covariates.
individual_effects	A list of 2 by N_individual_parameters matrix of slopes. The list runs over layers/timesteps. In each element, the first row gives effects of focal characteristics (on out-degree). The second row gives effects of target characteristics (on in-degree).
dyadic_effects	A list of N_dyadic_parameters vectors of slopes.

Value

A list of objects including: network (an adjacency tensor of binary outcomes), tie_strength (an adjacency tensor with probability weights), group_ids (a vector of length N_id, giving the group of each individual), individual_predictors (the supplied covariate data is saved along with the network data), and dyadic_predictors (the supplied covariate data is saved along with the network data).

ctross/STRAND documentation built on Dec. 15, 2024, 6:02 a.m.