fit_p1: Fitting function for p1 models

In rugbel/p2model: Maximum Likelihood Estimation of p2 models

Description

Estimates a p1 model via (exact) maximum likelihood.

Usage

fit_p1(y, XnS, XnR, XvD, XvC, trace = FALSE, init = NULL,
  opt = nlminb, ...)

Arguments

y	An adjacency matrix of size g x g.
XnS	Matrix of sender effects of size g x ks.
XnR	Matrix of receiver effects of size g x kr.
XvD	Density effects, a 3-dim array of size g x g x kd.
XvC	Reciprocity effects, a 3-dim array of size g x g x kc.
trace	TRUE for tracing information during the estimation. Default is FALSE.
init	Optional starting value for model parameters. Default is NULL.
opt	Name of the optimising function. Default is nlminb.
...	Optional arguments passed to the optimiser.

Details

The function is included for compatibility with fit_p2, and it might also be useful to obtain starting values.

Value

The returned value is an object of class "p1", a list containing the following components:

theta	the vector of model estimates.
loglik	the log likelihood function at the estimate.
AIC, BIC	model selection criteria at the estimate.
XnS.null, XnR.null	logical, flagging whether there are no sender or receiver effects, respectively.
seed	the random seed used for estimation (when M>0).
theta.cov	Variance matrix of estimates.
theta.se	Standard errors of estimates.
opt	Optimiser employed for estimation.
opt.details	Object returned by the optimiser.
ADobj	Object returned by TMB:MakeADFun, containing the log likelihood function and its gradient (among its components).
model.data	List containing all the model data, fed to TMB:MakeADFun.
sdrep	Summary object returned by TMB::sdreport. This is an object with many slots, useful for TMB users.

Author(s)

Ruggero Bellio

Examples

# Analysis of the  kracknets data from the NetData package
library(NetData)
data(kracknets)
# data preparation
g <- 21
Y <- matrix(0, g, g)
ind <-1
for(i in 1:nrow(friendship_data_frame)){
   sele <- friendship_data_frame[i, ]
   Y[sele$ego, sele$alter] <- sele$friendship_tie
   }
Xn <- model.matrix(~ AGE + TENURE, attributes)[, -1]
XvD <- array(1, dim=c(g, g, 4))
for(i in 1:g)
 for(j in 1:g){
    XvD[i, j, 2] <- as.numeric(attributes$DEPT[i]==attributes$DEPT[j])
    XvD[i, j, 3] <- as.numeric(attributes$LEVEL[i]==attributes$LEVEL[j])
    XvD[i, j, 4] <- abs(attributes$AGE[i] - attributes$AGE[j])
  }
XvC <- array(1, dim=c(g, g, 1))
# Now we are ready to fit the model
mod <- fit_p1(Y, Xn, Xn, XvD, XvC)
summary(mod)

rugbel/p2model documentation built on March 8, 2021, 8:16 p.m.