Query on getting back association parameter from data simulation in weibull-ph-agh

[faithghlee]

Hello Dr Rizopoulos,

I am simulating data coming from an Exponential PH model and I am trying to feed it to the jointModel() function to see whether I get back my parameter estimates.

You have mentioned in page 95 of your joint modelling book that survreg fits AFT model and thus if we want MLE's under the relative risk parameterization we will have to supply relevant initiation of the values to the jointModel() function.

In specifying the "weibull-PH-aGH" , I noticed that my association parameter was extremely far from what I had simulated (ie. -0.0006 vs 0.2). However, without using survreg (using coxph) and calling jointModel directly (without init), I was able to get back my association parameter (0.1957).

I am wondering what did I do wrong or if theres something I should be aware of?

fitLME<- lme(measurements ~ Time + FixedCovariate, random= ~ Time|ID, data = longitudinalDataframe)

marginal.survJM <- survreg(Surv(times, status)~W1+W2+X2, data=survivalDF, dist ="exponential", x=TRUE) #this is the Exponential AFT fit 

### CONVERT PARAMETERS AS PER PAGE 95 of Dr Rizopoulos' book. 

init.list<- list(betas = fixef(fitLME), sigma = fitLME$sigma, D = getVarCov(fitLME), 
                 gammas = -coef(marginal.survJM)/marginal.survJM$scale,
                 sigma.t = 1)
JMFITWB<- jointModel(fitLME, marginal.survJM, timeVar= "Time", scaleWB=1, init = init.list, method="weibull-PH-aGH")
summary(JMFITWB)

coxPhModel <- coxph(Surv(times, status)~W1+W2+X2, data=survivalDF, x=TRUE)



JMFITCOX <- jointModel(fitLME, coxPhModel, timeVar = "Time", scaleWB=1)
summary(JMFITCOX)

Really would appreciate any help or guidance.

Thank you,
Faith