CODEWITHSUNDEEP
rsurvivalnon-standard-evaluation
eac2222
eac2222

Reputation: 191

variable variable name in argument to `tdc()` in `survival::tmerge()`

Using tmerge() to code time-varying covariate for survival: What if I don't know what covariate I want to code until runtime?

Pretend data: two people, and two time-varying covariates: bmi, and modifiedBmi.

personWise <- data.frame(personId=1:2, startDay=0, endDay=1:2 * 300, 
  event=c(TRUE, FALSE))
print.data.frame(personWise)

  personId startDay endDay event
1        1        0    300  TRUE
2        2        0    600 FALSE

visitWise <- data.frame(
  personId=c(1, 1, 1, 2, 2),
  visitDay=c(50, 150, 200, 20, 500),
  bmi=c(3, 4, 3, 6, 5),
  modifiedBmi=c(1, 7, 8, 2, 6)
)
print.data.frame(visitWise)

  personId visitDay bmi modifiedBmi
1        1       50   3           1
2        1      150   4           7
3        1      200   3           8
4        2       20   6           2
5        2      500   5           6

If I want to code bmi as my time-varying covariate, I can use tmerge() like this:

tv <- tmerge(data1=personWise, data2=personWise, id=personId, tstart=startDay, 
  tstop=endDay, event=event(endDay, event))
tv <- tmerge(data1=tv, data2=visitWise, id=personId, 
    bmiMeasure=tdc(visitDay, bmi))

Or, if I'm interested in modifiedBmi, I could change that second tv() call to include . . . bmiMeasure=tdc(visitDay, modifiedBmi))

But what if I want to run a number of models, possibly on different datasets, changing which BMI measure I care about programatically?

Ideally I want something like bmiCovariate <- "bmi" and . . . bmiMeasure=tdc(visitDay, bmiCovariate)), but of course that doesn't work because the arguments to tdc() are getting some version of nonstandard evaluation?

Is there a way to do this?

Upvotes: 0

Views: 41

Answers (1)

xilliam
xilliam

Reputation: 2259

Perhaps you could set up control logic that determines which measure is used?

bmiCovariate <- "bmi"

if(bmiCovariate == "bmi"){
  
  tv <- tmerge(data1=tv, data2=visitWise, id=personId, 
               bmiMeasure=tdc(visitDay, bmi))
  
} else {   
  
  tv <- tmerge(data1=tv, data2=visitWise, id=personId, 
               bmiMeasure=tdc(visitDay, modifiedBmi))
  
}

Upvotes: 1

Related Questions