dynInfo: Dynamic Information of an Extra Longitudinal Measurement

Description

Using the available longitudinal information up to a particular time point, this function computes an estimate of the information we again by obtaining an extra longitudinal measurement.

Usage

dynInfo(object, newdata, Dt, K = 5, M = 500, idVar = "id", 
    simulateFun = function (eta, scale) rnorm(length(eta), eta, scale), 
    seed = 1L)

Arguments

object

an object inheriting from class JMBayes.

newdata

a data frame that contains the longitudinal and covariate information for the subject for whom we wish to plan the next measurement. The names of the variables in this data frame must be the same as in the data frames that were used to fit the linear mixed effects model (using lme()) and the survival model (using coxph() that were supplied as the two first argument of jointModelBayes. In addition, this data frame should contain a variable that identifies the subject (see also argument idVar).

numeric scalar denoting the length of the time interval to search for the optimal time point of the next measurement, i.e., the interval is $(t, t + Delta t]$ with $Delta t$ given by Dt.

numeric scalar denoting the number of time points to cosider in the interval $(t, t + Delta t]$.

idVar

the name of the variable in newdata that identifies the subject.

simulateFun

a function based on which longitudinal measurement can be simulated. This should have as a main argument the variable eta that denotes the subject-specific linear predictor from the mixed model, and possibly a scale parameter.

a numeric scalar denoting the number of Monte Carlo samples.

seed

a numeric scalar

Value

A list with components:

summary

a numeric matrix with first column the time points at which the longitudinal measurement is hypothetically taken, second column the estimated information we gain by obtaining the measurement, and third column the estimated cumulative risk of an event up to the particular time point denoted in the first column.

full.results

a numeric matrix with columns representing the time points, rows the Monte Carlo samples, and entries the value of log posterio predictive density.

Details

This functions computes the following posterior predictive distribution $$E_{Y} [ E_{T^* | Y} (\log p (T_j^* \mid T_j^* > u, \{ Y_j(t), y_j(u) \}, D_n \bigr )) ],$$ where $T_j^*$ denotes the time-to-event for subject $j$ for whom we wish to plan the next visit, $Y_j(t)$ the available longitudinal measurements of this subject up to time $t$, $y_j(u)$ the future longitudinal measurement we wish to plan at time $u > t$, and $D_n$ the data set that was used to fit the joint model.

Examples

Run this code

# NOT RUN {
# we construct the composite event indicator (transplantation or death)
pbc2$status2 <- as.numeric(pbc2$status != "alive")
pbc2.id$status2 <- as.numeric(pbc2.id$status != "alive")

# we fit the joint model using splines for the subject-specific 
# longitudinal trajectories and a spline-approximated baseline
# risk function
lmeFit <- lme(log(serBilir) ~ ns(year, 3),
    random = list(id = pdDiag(form = ~ ns(year, 3))), data = pbc2)
survFit <- coxph(Surv(years, status2) ~ drug, data = pbc2.id, x = TRUE)
jointFit <- jointModelBayes(lmeFit, survFit, timeVar = "year")

dynInfo(jointFit, newdata = pbc2[pbc2$id == 2, ], Dt = 5)[[1]]
# }