remulateTieEffects: Remulate Tie Effects

In remulate: Simulate Dynamic Networks from Relational Event Models

Remulate Tie Effects

Description

This page lists the effects that are available in the remulate package for the tie oriented relational event model.

Usage

remulateTieEffects(endogenous = NULL)

Arguments

endogenous

Logical or NULL. If 'TRUE', returns only endogenous effects. If 'FALSE', returns only exogenous effects. If 'NULL' (default), returns both endogenous and exogenous effects.

Details

The attr_actors object for exogenous effects based on actor covariates (send, receive, same, difference, average, max, min) contains at least three columns (actor,time,attribute). It should be constructed as follows: Each row refers to the attribute value of actor i at timepoint t. The first column contains the actor names (corresponding to the vector of names in the actors argument of remulateTie). The second column contains the time when attr_actors change (set to zero if the attr_actors do not vary over time). At least one of the subsequent columns must contain values for the attr_actors with column name corresponding to variable name specified in the effect specification.

The attribute object for exogenous effect dyad contains at least three columns (sender_id,receiver_id,attribute). It should be constructed as follows: First column must contain sender id, second column receiver id, at least one of the subsequent columns must contain values for the attr_actors with column name corresponding to variable name specified in the effect specification.

if param is a data frame, it must have three columns: sender, receiver, and value (numeric), representing the parameter value for thay dyadic pair. The data.frame must contain all pairs of actors or dyads corresponding to the riskset.

if param is a function, it's first argument must be 't', corresponding to the time. The function may have additional arguments.

The indices aregument in the interact effect corresponds to the position of the specified effects in the effects argument of remulateTie for which the interaction needs to be computed. The individual constitutive effects for an interaction must be specified before the interact term in the effects argument. To omit the individual constitutive effects in the generation, specify the param arugment to zero.

Value

Returns a character vector of available effects for the effects argument for the function remulateTie.

Remulate Effects

baseline

Baseline tendency for dyads to create events. The statistic equals to 1 for all dyads (i,j) in the riskset. The parameter for baseline controls the average number of events per unit time.

Endogenous effects (Dyad statistics):

inertia

Inertia is the tendency to create an event i->j if the event i->j occurred in the past. The statistic at timepoint t for dyad (i,j) is equal to the number of (i,j) events before timepoint t. Note: if scaling is "prop" for inertia, the statistic for dyad (i,j) at time t is divided by the out degree of the sender i at time t.

reciprocity

Reciprocity is the tendency to create an event i->j if j->i occurred in the past.The statistic at timepoint t for dyad (i,j) is equal to the number of (j,i) events before timepoint t. Note: if scaling is "prop" for reciprocity, the statistic for dyad (i,j) at time t is divided by the in degree of the sender i at time t.

tie

Tie effect is the tendency to create an event i->j if the event i->j occurred at least once in the past. The statistic at timepoint t for dyad (i,j) is equal to 1 if a an event i->j occurred before timepoint t

Endogenous effects (Triadic statistics):

otp

Outgoing Two Path effect is the tendency to create an event i->j if they have past outgoing two-paths between them (i->h->j). The statistic for dyad (i,j) at timepoint t is equal to the minimum of past (i,h), (h,j) events, summed over all h.

itp

Incoming Two Path effect is the tendency to create an event i->j if they have past incoming two-paths between them (i<-h<-j). The statistic for dyad (i,j) at timepoint t is equal to the minimum of past (j,h), (h,i) events, summed over all h.

osp

Outgoing Shared Partners effect is the tendency to create an event i->j if they have past outgoing shared partners between them (i->h<-j). The statistic for dyad (i,j) at timepoint t is equal to the minimum of past (i,h), (j,h) events, summed over all h.

isp

Incoming Shared Partners effect is the tendency to create an event i->j if they have past incoming shared partners between them (i<-h->j). The statistic for dyad (i,j) at timepoint t is equal to the minimum of past (h,i), (h,j) events, summed over all h.

Endogenous effects (Node statistics):

indegreeSender

In degree effect of the sender is the tendency to create an event i->j if i has received more events in the past. The statistic at timepoint t for dyad (i,j) is equal to the number of events received by actor i before timepoint t. Note: if scaling is "prop" for indegreeSender, the statistic for dyad (i,j) at time t is divided by the number of past events until time t.

indegreeReceiver

In degree effect of receiver is the tendency to create an event i->j if j has received more events in the past. The statistic at timepoint t for dyad (i,j) is equal to the number of events received by actor j before timepoint t. Note: if scaling is "prop" for indegreeReceiver, the statistic for dyad (i,j) at time t is divided by the number of past events until time t.

outdegreeSender

Out degree effect of sender is the tendency to create an event i->j if i has sent more events in the past. Note: if scaling is "prop" for outdegreeSender, the statistic for dyad (i,j) at time t is divided by by the number of past events until time t.

outdegreeReceiver

Out degree effect of receiver is the tendency to create an event i->j if j has sent more events in the past. Note: if scaling is "prop" for outdegreeReceiver, the statistic for dyad (i,j) at time t is divided by the number of past events until time t.

totaldegreeSender

Total degree effect of sender is the tendency to create an event i->j if i has sent and received more events in the past.

totaldegreeReceiver

Total degree effect of receiver is the tendency to create an event i->j if j has sent and received more events in the past.

Endogenous effects (Participating Shifts):

psABBA

AB-BA Pacticipating shift (turn receiving) is the tendency to create an event j->i at timepoint t if event i->j occurred at timepoint t-1. The psABBA statistic is equal to one for the dyad (j.i) that will create the participation shift at timepoint t.

psABBY

AB-BY Participating shift (turn receiving) is the tendency to create an event j->h at timepoint t if event i->j occurred at timepoint t-1. The psABBY statistic is equal to one for the dyads (j,h) for all h not equal to i, that will create the participation shift at timepoint t.

PSABAY

AB-AY Participating shifts (turn continuing) is the tendency to create an event i->h at timepoint t if event i->j occurred at timepoint t-1. The PSABAY statistic is equal to one for the dyads (i,h) for all h not equal to j, that will create the participation shift at timepoint t.

psABXA

AB-XA Participating shifts (turn usurping) is the tendency to create an event h->i at timepoint t if event i->j occurred at timepoint t-1. The psABXA statistic is equal to one for the dyads (h,i) for all h not equal to j, that will create the participation shift at timepoint t.

psABXB

AB-XB Participating shifts (turn usurping) is the tendency to create an event h->j at timepoint t if event i->j occurred at timepoint t-1. The psABXB statistic is equal to one for the dyads (h,j) for all h not equal to i, that will create the participation shift at timepoint t.

psABXY

AB-XY Participating shifts (turn usurping) is the tendency to create an event h->k at timepoint t if event i->j occurred at timepoint t-1. The psABXB statistic is equal to one for the dyads (h,k) for all h not equal to i and k not equal to j, that will create the participation shift at timepoint t.

Endogenous effects (Recency statistics):

recencyContinue

The recencyContinue effect refers to a recency statistic similar to what is described in Vu et al. (2017) and Mulder and Leenders (2019). For each timepoint t, for directed dyad (i,j) the statistic is equal to 1/(the time that has past since the dyad was last active + 1).

recencySendSender

The recencySendSender effect refers to a recency statistic similar to what is described in Vu et al. (2017) and Mulder and Leenders (2019). For each timepoint t, for directed dyad (i,j) the statistic is equal to 1/(the time that has past since sender i was last active as sender + 1).

recencySendReceiver

The recencySendReceiver effect refers to a recency statistic similar to what is described in Vu et al. (2017) and Mulder and Leenders (2019). For each timepoint t, for directed dyad (i,j) the statistic is equal to 1/(the time that has past since receiver j was last active as sender + 1).

recencyReceiveSender

The recencyReceiveSender effect refers to a recency statistic similar to what is described in Vu et al. (2017) and Mulder and Leenders (2019). For each timepoint t, for directed dyad (i,j) the statistic is equal to 1/(the time that has past since sender i was last active as receiver + 1).

recencyReceiveReceiver

The recencyReceiveReceiver effect refers to a recency statistic similar to what is described in Vu et al. (2017) and Mulder and Leenders (2019). For each timepoint t, for directed dyad (i,j) the statistic is equal to 1/(the time that has past since receiver j was last active as receiver + 1).

Exogenous effects (Node attr_actors):

send

Sender covariate: The tendency to create an event i->j when i has a high attribute value.

receive

Receiver covariate: The tendency to create an event i->j when j has a high attribute value.

Exogenous effects (Tie Attribute):

dyad

dyad attribute value is the tendency to create an event i -> j when (i,j) has a high attribute value.

same

(Homophily) is the tendency to create an event i->j if actors i and j have the same attribute values

difference

(Heterophily) is the tendency to create an event i->j if actors i and j have a high absolute difference in attribute values

average

average attribute value for dyad (i,j) is the average of the attribute values for actors i, j

minimum

minimum attribute value for dyad (i,j) is the smaller of the attribute values for actors i , j

maximum

maximum attribute value for dyad (i,j) is the bigger of the attribute values for actors i , j

Examples

#To specify an endogenous effect (example: inertia)

effects <- ~ inertia(0.1, scaling = "std")

#To specify an exogenous effect (example: same)

cov <- data.frame(
  actor = 1:10,
  time = rep(0, 10),
  gender = sample(c(0, 1), replace = TRUE, 10),
  age = sample(20:30, 10, replace = TRUE)
)

effects <- ~ same(0.2, variable = "gender", attr_actors = cov)

#To specify an exogenous dyadic effect (example: dyad)

cov <- expand.grid(1:10, 1:10)
cov <- cov[cov[, 1] != cov[, 2], ]
cov$grade <- runif(90, 1, 10)

effects <- ~ dyad(0.2, variable = "grade", attr_actors = cov)

#If parameter is constant

effects <- ~ inertia(0.3) + 
  same(0.2, variable = "gender", attr_actors = cov) + 
  reciprocity(-0.1) + 
  itp(0.01)

#If parameter varies with time

effects <- ~ inertia(param = function(t) exp(-t)) + 
  same(0.2, variable = "gender", attr_actors = cov) +
  reciprocity(-0.1) + 
  itp(0.01)

 #If parameter varies across dyads or actors
 rs <- expand.grid(1:10,1:10)
 rs <- rs[rs[,1] != rs[, 2],]

 param_df <- as.data.frame(rs)
 param_df$beta = runif(nrow(rs),-0.1,0.1)

 effects <- ~ remulate::baseline(-3)+
    remulate::inertia(param_df) +
    remulate::reciprocity(0.1)

 
#To specify an interaction (example: between inertia and same constitutive effects)

effects <- ~ inertia(0.3) + 
  same(0.2, variable = "gender", attr_actors = cov) + 
  reciprocity(-0.1) + 
  itp(0.01) + 
  interact(0.1, indices = c(1, 2))

remulate documentation built on April 16, 2025, 5:09 p.m.