Hist: Create an event history response variable

Description

Functionality for managing censored event history response data. The function can be used as the left hand side of a formula: Hist serves prodlim in a similar way as Surv from the survival package serves `survfit'. Hist provides the suitable extensions for dealing with right censored and interval censored data from competing risks and other multi state models. Objects generated with Hist have a print and a plot method.

Usage

Hist(
  time,
  event,
  entry = NULL,
  id = NULL,
  cens.code = "0",
  addInitialState = FALSE
)

Value

An object of class Hist for which there are print and plot methods. The object's internal is a matrix with some of the following columns:

time: the right censored times
L: the left endpoints of internal censored event times
R: the right endpoints of internal censored event times
status: 0 for right censored, 1 for exact, and 2 for interval censored event times.
event: an integer valued numeric vector that codes the events.
from: an integer valued numeric vector that codes the from states of a transition in a multi state model.
to: an integer valued numeric vector that codes the to states of a transition in a multi state model.

Further information is stored in attributes. The key to the official names given to the events and the from and to states is stored in an attribute "states".

Arguments

time: for right censored data a numeric vector of event times -- for interval censored data a list or a data.frame providing two numeric vectors the left and right endpoints of the intervals. See Details.
event: A vector or a factor that specifies the events that occurred at the corresponding value of time. Numeric, character and logical values are recognized. It can also be a list or a data.frame for the longitudinal form of storing the data of a multi state model -- see Details.
entry: Vector of delayed entry times (left-truncation) or list of two times when the entry time is interval censored.
id: Identifies the subjects to which multiple events belong for the longitudinal form of storing the data of a multi state model -- see Details.
cens.code: A character or numeric vector to identify the right censored observations in the values of event. Defaults to "0" which is equivalent to 0.
addInitialState: If TRUE, an initial state is added to all ids for the longitudinal input form of a multi-state model.

Author

Thomas A. Gerds tag@biostat.ku.dk, Arthur Allignol arthur.allignol@fdm.uni-freiburg.de

Details

*Specification of the event times*

If time is a numeric vector then the values are interpreted as right censored event times, ie as the minimum of the event times and the censoring times.

If time is a list with two elements or data frame with two numeric columns The first element (column) is used as the left endpoints of interval censored observations and the second as the corresponding right endpoints. When the two endpoints are equal, then this observation is treated as an exact uncensored observation of the event time. If the value of the right interval endpoint is either NA or Inf, then this observation is treated as a right censored observation. Right censored observations can also be specified by setting the value of event to cens.code. This latter specification of right censored event times overwrites the former: if event equals cens.code the observation is treated as right censored no matter what the value of the right interval endpoint is.

*Specification of the events*

If event is a numeric, character or logical vector then the order of the attribute "state" given to the value of Hist is determined by the order in which the values appear. If it is a factor then the order from the levels of the factor is used instead.

**Normal form of a multi state model**

If event is a list or a data.frame with exactly two elements, then these describe the transitions in a multi state model that occurred at the corresponding time as follows: The values of the first element are interpreted as the from states of the transition and values of the second as the corresponding to states.

**Longitudinal form of a multi state model**

If id is given then event must be a vector. In this case two subsequent values of event belonging to the same value of id are treated as the from and to states of the transitions.

Examples

Run this code



## Right censored responses of a two state model
## ---------------------------------------------

Hist(time=1:10,event=c(0,1,0,0,0,1,0,1,0,0))

## change the code for events and censored observations

Hist(time=1:10,event=c(99,"event",99,99,99,"event",99,"event",99,99),cens.code=99)

TwoStateFrame <- SimSurv(10)
SurvHist <- with(TwoStateFrame,Hist(time,status))
summary(SurvHist)
plot(SurvHist)

## Right censored data from a competing risk model
## --------------------------------------------------

CompRiskFrame <- data.frame(time=1:10,event=c(1,2,0,3,0,1,2,1,2,1))
CRHist <- with(CompRiskFrame,Hist(time,event))
summary(CRHist)
plot(CRHist)

## Interval censored data from a survival model
icensFrame <- data.frame(L=c(1,1,3,4,6),R=c(2,NA,3,6,9),event=c(1,1,1,2,2))
with(icensFrame,Hist(time=list(L,R)))

## Interval censored data from a competing risk model
with(icensFrame,Hist(time=list(L,R),event))

## Multi state model
MultiStateFrame <- data.frame(time=1:10,
			      from=c(1,1,3,1,2,4,1,1,2,1),
			      to=c(2,3,1,2,4,2,3,2,4,4))
with(MultiStateFrame,Hist(time,event=list(from,to)))

##  MultiState with right censored observations

MultiStateFrame1 <- data.frame(time=1:10,
			      from=c(1,1,3,2,1,4,1,1,3,1),
			      to=c(2,3,1,0,2,2,3,2,0,4))
with(MultiStateFrame1,Hist(time,event=list(from,to)))

## Using the longitudinal input method 
MultiStateFrame2 <- data.frame(time=c(0,1,2,3,4,0,1,2,0,1),
			      event=c(1,2,3,0,1,2,4,2,1,2),
			      id=c(1,1,1,1,2,2,2,2,3,3))
 with(MultiStateFrame2,Hist(time,event=event,id=id))

Run the code above in your browser using DataLab