I’m attempting to estimate the average treatment effect of a baseline treatment on a survival outcome by fitting a doubly robust estimator using the ate() function in the package riskRegression. Documentation states that the argument treatment must be either the name of the treatment variable or, for IPTW, a glm object from a logistic regression (treatment regressed on covariates). When I do the latter, I get the error:

`Processing

• Prepare influence function: outcome censoring done
• Point estimation:Error: Not a matrix.`

I do not get this error when using the name of the treatment variable instead. Given that I am trying to use a doubly robust estimator, however, I want to use IPTW but I cannot figure out what is causing this error when I specify the glm object.

Apologies for not having a reproducible example, but here is the relevant code below, which closely follows the package documentation examples for the ate function. The scenario is for baseline (time-invariant) treatment and covariates with a survival outcome.

m.event <- coxph(Surv(fu_time, event) ~ treatment + race + sex, data=compdat, x=TRUE)
m.censor <- coxph(Surv(fu_time, event==0) ~ treatment + race + sex, data=compdat, x=TRUE)
m.treatment <- glm(treatment~race+sex, data=compdat, family=binomial(link="logit"))  

ateRobust <- ate(event = m.event,
                   censor=m.censor,
                   treatment=m.treatment,
                   data=compdat,
                   times=0)