censordata: Censor the data to simulate a particular pattern of dropouts

In rchendrickson/pmsimstats: Personalized medicine simulated statistics

Censor the data to simulate a particular pattern of dropouts

Description

censordata simulates a pattern of participant dropouts

Usage

censordata(dat, trialdesign, censorparam)

Arguments

dat	dat file as produced by generateData
trialdesign	a trial design as defined by buildtrialdesign
censorparam	a data.table provides 3 named parameters: (1) beta0 - the total rate (approximately) of censoring (0-1) (2) beta1 - the proportion of the total dropout that should be biased (0-1) (3) eb1 - the exponent used to define the shape of the curve defining biased dropout. See example, below.

Details

This function inputs a dat file as would be produced by generateData and selects a subset to drop out (censor). The probability of a particular timepoint being selected for censoring in the initial pass is the sum of two factors: one that censors at a set rate as a function of time, and one that biases the rate of censoring based on the trajectory of symptoms, with a large improvement in symptoms from baseline decreasing the probability of dropout, while a small change or a worsening increases the probability of dropout. In a separate, second pass all dropouts are carried forward, such that once a datapoint has been censored, all future datapoints for that particular participant are also censored

Value

A new data file, with the censored data points replaced by NA, will be returned

Examples

# Create a set of censoring params to
# feed to censordata one at a time:
censorparams<-data.table(
  names=c("balanced","more of flat","more of biased","high dropout"),
  beta0=c(.05,.05 ,.05,.15),
  beta1=c(.5,.2,.8 ,.5),
  eb1=  c(2,  2  ,2  ,2 )
  )

rchendrickson/pmsimstats documentation built on Nov. 28, 2024, 11:05 a.m.