The list of methods that apply to Surv objects
# S3 method for Surv
anyDuplicated(x, ...)
# S3 method for Surv
as.character(x, ...)
# S3 method for Surv
as.data.frame(x, ...)
# S3 method for Surv
as.matrix(x, ...)
# S3 method for Surv
c(...)
# S3 method for Surv
duplicated(x, ...)
# S3 method for Surv
format(x, ...)
# S3 method for Surv
head(x, ...)
# S3 method for Surv
is.na(x)
# S3 method for Surv
length(x)
# S3 method for Surv
mean(x, ...)
# S3 method for Surv
median(x, na.rm=FALSE, ...)
# S3 method for Surv
names(x)
# S3 method for Surv
names(x) <- value
# S3 method for Surv
quantile(x, probs, na.rm=FALSE, ...)
# S3 method for Surv
plot(x, ...)
# S3 method for Surv
rep(x, ...)
# S3 method for Surv
rep.int(x, ...)
# S3 method for Surv
rep_len(x, ...)
# S3 method for Surv
rev(x)
# S3 method for Surv
t(x)
# S3 method for Surv
tail(x, ...)
# S3 method for Surv
unique(x, ...)
a Surv object
a vector of probabilities
remove missing values from the calculation
a character vector of up to the same length as x, or
NULL
other arguments to the method
These functions extend the standard methods to Surv objects.
(There is no central index of R methods, so there may well be useful
candidates that the author has missed.)
The arguments and results from these are mostly as expected, with the
following further details:
The as.character function uses "5+" for right censored
at time 5, "5-" for left censored at time 5, "[2,7]" for an
observation that was interval censored between 2 and 7,
"(1,6]" for a counting process data denoting an observation which
was at risk from time 1 to 6, with an event at time 6, and
"(1,6+]" for an observation over the same interval but not ending
with and event.
For a multi-state survival object the type of event is appended to
the event time using ":type".
The print and format methods make use of
as.character.
The length of a Surv object is the number of
survival times it contains, not the number of items required to
encode it, e.g., x <- Surv(1:4, 5:8, c(1,0,1,0)); length(x)
has a value of 4.
Likewise names(x) will be NULL or a vector of length 4.
(For technical reasons, any names are actually stored in the
rownames attribute of the object.)
For a multi-state survival object levels returns the
names of the endpoints, otherwise it is NULL.
The median, quantile and plot methods
first construct a survival curve using survfit, then apply
the appropriate method to that curve.
The xtfrm method, which underlies sort and order,
sorts by time, with censored after uncensored within a tied time.
For an interval censored observation the midpoint is used.
For (time1, time2) counting process data, sorting is by time2, censoring,
and then time1.
The unique method treats censored and uncensored
observations at the same time as distinct, it returns a Surv object.
The concatonation method c() is asymmetric, its first
argument determines the execution path. For instance
c(Surv(1:4), Surv(5:6)) will return a Surv object of length 6,
c(Surv(1:4), 5:6) will give an error, and
c(5:6, Surv(1:4)) is equivalent to
c(5:6, as.vector(Surv(1:4))) which is a numeric of length 10.
Surv