library(survival) # # A Cox model with a factor, followed by survexp. # pfit2 <- coxph(Surv(time, status > 0) ~ trt + log(bili) + log(protime) + age + platelet + sex, data = pbc) esurv <- survexp(~ trt, ratetable = pfit2, data = pbc) temp <- pbc temp$sex2 <- factor(as.numeric(pbc$sex), levels=2:0, labels=c("f", "m", "unknown")) esurv2 <- survexp(~ trt, ratetable = pfit2, data = temp, rmap=list(sex=sex2)) # The call components won't match, which happen to be first all.equal(unclass(esurv)[-1], unclass(esurv2)[-1]) # Check that the ratetableDate function is okay # Datedate <- function(x) { # Dates have an origin of 1/1/1970, dates of 1/1/1960 offset <- as.numeric(as.Date("1970-01-01") - as.Date("1960-01-01")) y <- as.numeric(x) + offset class(y) <- "date" y } as.data.frame.date <- as.data.frame.vector # needed to make the functions work n <- nrow(lung) tdata <- data.frame(age=lung$age + (1:n)/365.25, sex = c('male', 'female')[lung$sex], ph.ecog = lung$ph.ecog, time = lung$time*3, status = lung$status, entry = as.Date("1940/01/01") + (n:1)*50) tdata$entry2 <- as.POSIXct(tdata$entry) tdata$entry3 <- Datedate(tdata$entry) p1 <- pyears(Surv(time, status) ~ ph.ecog, data=tdata, ratetable=survexp.us, rmap= list(age=age*365.25, sex=sex, year=entry)) p2 <- pyears(Surv(time, status) ~ ph.ecog, data=tdata, ratetable=survexp.us, rmap= list(age=age*365.25, sex=sex, year=entry2)) p3 <- pyears(Surv(time, status) ~ ph.ecog, data=tdata, ratetable=survexp.us, rmap= list(age=age*365.25, sex=sex, year=entry3)) all.equal(p1$expected, p2$expected) all.equal(p1$expected, p3$expected) # Now a ratetable with ordinary dates rather than US census style year trate <- survexp.us attr(trate, 'type') <- c(2,1,3) p4 <- pyears(Surv(time, status) ~ ph.ecog, data=tdata, ratetable=trate, rmap= list(age=age*365.25, sex=sex, year=entry)) p5 <- pyears(Surv(time, status) ~ ph.ecog, data=tdata, ratetable=trate, rmap= list(age=age*365.25, sex=sex, year=entry2)) p6 <- pyears(Surv(time, status) ~ ph.ecog, data=tdata, ratetable=trate, rmap= list(age=age*365.25, sex=sex, year=entry3)) #all.equal(p1$expected, p4$expected) # this won't be true, US special is special all.equal(p4$expected, p5$expected) all.equal(p5$expected, p6$expected)