Skip to contents

Simulate a Relational Event History (REH)

generate_reh.Rd

This function simulates a sequence of relational events based on a specified statistical model derived from a relational event model (REM). It uses covariates, a list of parameters, and the number of events to generate a relational event history (REH) with associated statistics.

Usage

generate_reh(parameters, covar, M, directed = FALSE)

Arguments

parameters

A named list of parameters for the model. Each parameter represents a specific effect or term in the tie-oriented model (e.g., send_z1, difference_z2, inertia). The parameter names must align with those supported by remreg::generate_formula().

covar

A data frame of covariates containing at least the following columns:

  • name: Numeric or character identifiers for actors.

  • z1: A numeric covariate representing some attribute of the actors.

  • z2: A numeric covariate representing another attribute of the actors.

  • time: The starting time for the simulation (default is usually 0). Each row represents an actor in the network.

M

An integer specifying the number of events to simulate.

directed

A logical value indicating whether the network is directed (TRUE) or undirected (FALSE).

Value

A data frame representing the simulated event history. The data frame contains the following columns:

  • time: The time of each event.

  • actor1: The identifier of the first actor involved in the event.

  • actor2: The identifier of the second actor involved in the event.

Details

The generate_reh() function simulates a series of relational events using the specified model parameters and covariates. The simulation is based on a relational event model (REM), a statistical framework for modeling and analyzing dynamic interactions between actors over time.

It performs the following steps:

  1. Formula Generation: The function uses remreg::generate_formula() to create a formula for the tie_effects argument of remstats::remstats().

  2. Initialization: A dummy first event is created to initialize the simulation.

  3. Simulation Loop: For each event:

    • The model parameters are applied to compute the event rates.

    • A new event is sampled based on the computed rates.

    • The event is added to the event history, and the model is updated.

  4. Endogeneity Updates: The function computes endogeneity statistics for subsequent events using remstats::remstats().

Standardization: All effects that allow scaling are standardized by default (scaling = 'std'), ensuring consistent computation of statistics.

Note

This function is inspired by the work presented in the article:

Lakdawala, R., Mulder, J., & Leenders, R. (2024). Simulating Relational Event Histories–Why and How. arXiv preprint arXiv:2403.19329.

References

Lakdawala, R., Mulder, J., & Leenders, R. (2024). Simulating Relational Event Histories–Why and How. arXiv preprint arXiv:2403.19329.

See also

remstats::remstats() for calculating statistics for tie-oriented models. generate_formula() for generating the formula used in this function.

Examples

# Define parameters
parameters <- list(
  baseline = -5,
  send_z1 = 0.5,
  difference_z2 = 0.2,
  inertia = 0.3
)

# Create covariate data
covar <- data.frame(
  name = 1:50, time = 0,
  z1 = rnorm(n = 50, mean = 0, sd = 1),
  z2 = rnorm(n = 50, mean = 0, sd = 1)
)

# Simulate 10 events
events <- generate_reh(parameters, covar, M = 10, directed = TRUE)
#> 1 
2 
3 
4 
5 
6 
7 
8 
9 
10 

print(events$edgelist)
#>          time actor1 actor2
#> 1  0.02037744      3     37
#> 2  0.02038913      3     37
#> 3  0.02041367      3     37
#> 4  0.02043817      3     37
#> 5  0.02048893      3     37
#> 6  0.02066064      3     37
#> 7  0.02070345      3     37
#> 8  0.02071180      3     37
#> 9  0.02076789      3     37
#> 10 0.02078205      3     37