In timonelmer/dena: DENA - dynamic ego-centric network analysis

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

load(file = "../m_cmm_vignette.RData")


# and other packages that we need for estimation and visualization
library(survival)

This vignette aims to describe how one can analyze multistate frailty models in which one target category (e.g., interacting with one's partner and family members at the same time) consists of the combination of other categories.

# load the dena package with the data and functions
library(dena)

load the example data

rm(list = ls())

data(simdat4) # see the end of the script for simulation code

this dataset contains transition from the alone state to four different target states (1= family, 2= friend, 3= partner, 4= partner + family). This is a particular case of a multistate model, where still all categories are competing against each other, but one category (i.e., "partner + family") consists of the combination of two other categories.

The typevariable in the dataset describes these target states:

head(simdat4)

table(simdat4$type)

Estimating the multistate model

From a modeling perspective, there are no differences, compared to a standard application of the multistate model, only that now there is a combined category of target states, of which the coefficients need to be interpreted in the light of the data collection process.

m.cmm <- cmm(Surv(time, type) ~ Covariate1 + Covariate2 + (1 | id), dat = simdat4, verbose = T)

Here are the estimated fixed effects of the model:

m.cmm[[1]]

or as a plot of the fixed effects:

plot.cmm(m.cmm[[1]])

Dataset simulation

set.seed(921)
simdat4 <- frailtySurv::genfrail(N = 100, K = 50, 
                                 beta = c(log(1), -log(0.5)),
                                 frailty = "gamma", theta = 2, 
                                 censor.rate = 0,
                                 lambda_0=function(t, tau=4.6, C=0.01) (tau*(C*t)^tau)/t)

colnames(simdat4) <- c("id","rep","time","event","Covariate1","Covariate2")
simdat4$type <- factor(sample(c("partner", "family", "partner + family","friend"), nrow(simdat4), replace = T, prob = c(0.2,0.3,0.2,0.3)))

#save(simdat4, file = "../data/simdat4.RData")