Epi: mcutLexis – R documentation

Pricing

Become an expert in R — Interactive courses, Cheat Sheets, certificates and more!

Get Started for Free

Documentation

Epi

mcutLexis

Cut follow-up at multiple event dates and keep track of order of events

Description

A generalization of cutLexis to the case where different events may occur in any order (but at most once for each).

Usage

mcutLexis( L0, timescale = 1, wh,
              new.states = NULL,
        precursor.states = transient(L0),
              seq.states = TRUE,
              new.scales = NULL,
            ties.resolve = FALSE )

Arguments

`L0`	A Lexis object.
`timescale`	Which time scale do the variables in `L0[,wh]` refer to. Can be character or integer.
`wh`	Which variables contain the event dates. Character or integer vector
`new.states`	Names of the events forming new states. If `NULL` equal to the variable names from `wh`.
`precursor.states`	Which states are precursor states. See `cutLexis` for definition of precursor states.
`seq.states`	Should the sequence of events be kept track of? That is, should A-B be considered different from B-A. If `FALSE`, the state with combined preceding events A and B will be called A+B (alphabetically sorted). May also be supplied as character: `s` - sequence, keep track of sequence of states occupied (same as `TRUE`), `u` - unordered, keep track only of states visited (same as `FALSE`) or `l`, `c` - last or current state, only record the latest state visited. If given as character, only the first letter converted to lower case is used.
`new.scales`	Should we construct new time scales indicating the time since each of the event occurrences.
`ties.resolve`	Should tied event times be resolved by adding random noise to tied event dates. If `FALSE` the function will not accept that two events occur at the same time for a person (ties). If `TRUE` a random quantity in the range `c(-1,1)/100` will be added to all event times in all records with at least one tie. If `ties.resolve` is numeric a random quantity in the range `c(-1,1)*ties.resolve` will be added to all event times in all records with at least one tie.

Value

A Lexis object with extra states created by occurrence of a number of intermediate events.

Author(s)

Bendix Carstensen, http://bendixcarstensen.com

Examples

# A dataframe of times
set.seed(563248)
dd <- data.frame( id = 1:30,
                 doN = round(runif(30,-30, 0),1),
                 doE = round(runif(30,  0,20),1),
                 doX = round(runif(30, 50,60),1),
                 doD = round(runif(30, 50,60),1),
                 # these are the event times
                 doA = c(NA,21,NA,27,35,NA,52, 5,43,80,
                         NA,22,56,28,53,NA,51, 5,43,80,
                         NA,23,NA,33,51,NA,55, 5,43,80),
                 doB = c(NA,20,NA,53,27,NA, 5,52,34,83,
                         NA,20,23,37,35,NA,52, 8,33,NA,
                         25,NA,37,40,NA,NA,15,23,36,61) )

# set up a Lexis object with time from entry to death/exit
Lx <- Lexis( entry = list(time=doE,
                           age=doE-doN),
              exit = list(time=pmin(doX,doD)),
       exit.status = factor(doD<doX,labels=c("OK","D")),
              data = dd )
summary( Lx )

# cut the follow-up at dates doA and doB
L2 <- mcutLexis( Lx, "time", wh=c("doA","doB"),
                 new.states = c("A","B"),
           precursor.states = "OK",
                 seq.states = TRUE,
                 new.scales = c("tfA","tfB") )
summary( L2 )
L2

# show the states
boxes( L2, boxpos=list(x=c(10,60,50,90,50,90),
                       y=c(50,50,90,90,10,10)),
           scale.R=100, show.BE=TRUE, DR.sep=c(" (",")"))


L3 <- mcutLexis( Lx, "time", wh=c("doA","doB"),
                 new.states = c("A","B"),
           precursor.states = "OK",
                 seq.states = FALSE,
                 new.scales = c("tfA","tfB") )
summary( L3 )
boxes( L3, boxpos=list(x=c(10,50,50,90,50),
                       y=c(50,50,90,50,10)),
           show.R=FALSE, show.BE=TRUE )

Epi

Statistical Analysis in Epidemiology

v2.44

GPL-2

Authors

Bendix Carstensen [aut, cre], Martyn Plummer [aut], Esa Laara [ctb], Michael Hills [ctb]

Initial release

2021-02-28