I'm estimating accelerated failure time survival models with time-varying covariates and would like to compute the deviance residuals for my estimated models.
For fixed-time covariates this is easy to achieve with survival::survreg and stats::residuals. Here's an example taken from this webpage (the data is the ovarian dataset included in the survival package):
survregLogLogistic <- survreg(Surv(futime, fustat) ~ ecog.ps + rx,
data=ovarian, dist = "loglogistic")
devRes <- residuals(survregLogLogistic, type="deviance")
However, survival::survreg cannot handle time-varying covariates (see here for an example of how this type of data can be formatted for survival analysis in R). Because of this I need to use either flexsurv::flexsurvreg or eha::aftreg, as follows (without time-varying covariates but this is enough to reproduce the issue):
flexsurvregLogLogistic <- flexsurvreg(Surv(futime, fustat) ~ ecog.ps + rx,
data=ovarian, dist = "llogis")
The estimation works fine but the residuals command does not work (looking closely at the survival::residuals function using methods(residuals) one can see that there is no sub-function for either the flexsurv::flexsurvreg or the eha::aftreg functions). This is the output of the residuals function:
> residuals(flexsurvregLogLogistic, type="deviance")
NULL
And now for eha::aftreg:
aftregLogLogistic <- aftreg(Surv(futime, fustat) ~ ecog.ps + rx,
data=ovarian, dist = "loglogistic")
> residuals(aftregLogLogistic, type="deviance")
NULL
How can I compute deviance (or potentially martingale) residuals from either flexsurv::flexsurvreg or eha::aftreg objects?
