remulate (version 2.1.0)

remulateTie: Simulate Relational Event Data - Tie-Oriented model

Description

A function to simulate relational event data by sampling from a tie-oriented relational event model.

Usage

remulateTie(
  effects,
  actors,
  endTime,
  events = NULL,
  startTime = 0,
  initial = 0,
  riskset = NULL,
  memory = c("full", "window", "window_m", "decay"),
  memoryParam = NULL
)

Value

An object of class "remulateTie". A data.frame containing the simulated event sequence with columns (time, sender, receiver). The "remulateTie" object has the following attributes:

statistics

An array with dimensions M x D x P, where M is the number of events, D is the number of dyads in the risk set, and P is the number of computed statistics.

evls

A matrix containing the event list with columns (dyad, time), where dyad represents the index of the (sender, receiver) pair in the risk set.

actors

A data.frame mapping the actor names provided by the user to the integer IDs used in internal computations.

riskset

A data.frame with columns (sender, receiver) containing the risk set used for dyad indices in the computed statistics and event list.

initial

A A numeric or data.frame object representing the network initialization, which can be a number of random initialization events or a data.frame specifying pre-existing ties.

effects

A formula object specifying the effects included in the model.

density

A numeric value indicating the density of the generated network, defined as the number of observed ties divided by N*(N-1), where N is the number of actors.

Arguments

effects

A formula object specifying the statistics used to simulate the network or an object of class "remstimate" containing a fitted object.

actors

A numeric or character vector representing the actor names.

endTime

A numeric value specifying the end time up to which the network should be simulated.

events

[Optional] An integer specifying the maximum number of events to simulate.

startTime

[Optional] A numeric value (default = 0) indicating the time at which the simulation should start.

initial

[Optional] A numeric or data.frame object (default = 0) specifying how to initialize the network. If an integer is provided, it represents the number of random events to sample before beginning data generation. If a data.frame is provided with columns (time, sender, receiver), it serves as an edgelist of initial events, after which subsequent events are predicted.

riskset

[Optional] A matrix with columns (sender, receiver) defining a custom risk set.

memory

[Optional] A string (default = "full") specifying the memory type used for computing statistics. `"full"` uses the entire event history. `"window"` considers only events occurring within a specified time window. `"window_m"` considers only a specified number of most recent events. `"decay"` applies an exponential decay, where older events contribute less based on elapsed time.

memoryParam

[Optional] A numeric value (> 0) defining the memory parameter based on the selected memory type. `"window"` defines the length of the time window. `"window_m"` specifies the number of past events to consider. `"decay"` represents the half-life (i.e., time until an event's weight is reduced to half).

Details

If time is irrelevant and only a specific number of events are desired, set time to Inf. If both time and events are supplied then the function stops simulating whenever the first stop condition is met

If effects is a 'remstimate' object, then the params are extracted automatically. Furthermore if exogenous statistics are used, the data.frames corresponding to the 'attr_actors' argument of the 'remstats' formula must be loaded in the environment.

A list of available statistics. See remulateTieEffects for details:

• baseline(param)

• send()

• receive()

• dyad()

• same()

• difference()

• average()

• minimum()

• maximum()

• inertia()

• reciprocity()

• tie()

• indegreeSender()

• indegreeReceiver()

• outdegreeSender()

• outdegreeReceiver()

• totaldegreeSender()

• totaldegreeReceiver()

• otp()

• itp()

• osp()

• isp()

• psABBA()

• psABBY()

• psABXA()

• psABXB()

• psABXY()

• psABAY()

• recencyContinue()

• recencySendSender()

• recencySendReceiver()

• recencyReceiveSender()

• recencyReceiveReceiver()

• rrankSend()

• rrankReceive()

• interact()

References

Lakdawala, R., Mulder, J., & Leenders, R. (2025). *Simulating Relational Event Histories: Why and How*. arXiv:2403.19329.

Examples

 # To generate events up to time '50' in a network of 25 actors with 
 # 200 random initial events
 # Exogenous attributes data.frame

 cov <- data.frame(
   id = 1:25, 
   time = rep(0, 25), 
   sex = sample(c(0, 1), 25, replace = TRUE, prob = c(0.4, 0.6)), 
   age = sample(20:30, 25, replace = TRUE) 
 )

 #Effects specification
 effects <- ~ remulate::baseline(-5) + 
             remulate::inertia(0.01) + 
             remulate::reciprocity(-0.04) + 
             remulate::itp(0.01, scaling = "std") + 
             remulate::same(0.02, variable = "sex", attr_actors = cov) + 
             remulate::interact(0.01, indices = c(2, 5))
 # Calling remulateTie
 remulate::remulateTie(
   effects, 
   actors  = 1:25, 
   endTime = 50, 
   events  = 500, 
   initial = 200
 )

 # To predict events, given an edgelist of initial events
 initialREH <- data.frame(
   time = seq(0.5, 100, 0.5), 
   sender = sample(1:25, 200, TRUE), 
   receiver = sample(1:25, 200, TRUE)
 )

 remulate::remulateTie(
   effects, 
   actors  = 1:25, 
   endTime = 150, 
   events  = 500, 
   initial = initialREH
 )

 # Custom risk set
 rs <- as.matrix(expand.grid(1:25, 1:25))
 rs <- rs[rs[, 1] != rs[, 2], ]

 custom_rs <- rs[sample(1:90, 50), ]

 remulate::remulateTie(
   effects, 
   actors  = 1:25, 
   endTime = 150, 
   events  = 500, 
   riskset = custom_rs
 )