[R] Trying to speed up an if/else statement in simulations

[William Dunlap]

inline below

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com


> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf
> Of nqf
> Sent: Monday, June 18, 2012 10:30 AM
> To: r-help at r-project.org
> Subject: [R] Trying to speed up an if/else statement in simulations
> 
> Dear R-help,
> 
> I am trying to write a function to simulate datasets of size n which contain
> two time-to-event outcome variables with associated 'Event'/'Censored'
> indicator variables (flag1 and flag2 respectively). One of these indicator
> variables needs to be dependent on the other, so I am creating the first and
> trying to use this to create the second using an if/else statement.
> 
> My data structure needs to follow this algorithm (for each row of the data):
> If flag1=1 then flag2 should be 1 with probability 0.95 and zero otherwise
> Else if flag1=0 then flag2 should be 1 with probability 0.5 and zero
> otherwise
> 
> I can set up this example quite simply using if else statements, but this is
> incredibly inefficient when running thousands of datasets:
> data<-as.data.frame(rbinom(10,1,0.5))
> colnames(data)<-'flag1'
> for (i in 1:n) {
>   if (data$flag1[i]==1) {data$flag2[i]<-rbinom(1,1,0.95)} else
> {data$flag2[i]<-rbinom(1,1,0.5)}
>  }

Do you mean
   n <- 10
   data <- data.frame(flag1 = rbinom(n, 1, 0.5))
   for(i in seq_len(n)) {
      if (data$flag1[i]==1) {data$flag2[i]<-rbinom(1,1,0.95)} else {data$flag2[i]<-rbinom(1,1,0.5)}
  }
?

> 
> 
> I think to speed up the simulations I would be better changing to
> vectorisation and using something like:
> ifelse(data$flag1==1,rbinom(1,1,0.95),rbinom(1,1,0.5))
> but the rbinom statements here generate one value and repeat this draw for
> every element of flag2 that matches the 'if' statement on flag1.

Assuming that n is nrow(data), use
   data$flag2 <- ifelse(data$flag1==1, rbinom(n, 1, 0.95), rbinom(n, 1, 0.5))
so you get n independent draws from each distribution instead of 1.  I prefer
to factor out the common arguments as in
   data$flag2 <- rbinom(n, 1, ifelse(data$flag1==1, 0.95, 0.5))

   
> 
> Is there a way to assign flag2 to a new bernoulli draw for each subject in
> the data frame with flag1=1?
> 
> I hope my question is clear, and thank you in advance for your help.
> 
> Thanks,
> Natalie
> PhD student, Reading University
> 
> P.S. I am using R 2.12.1 on Windows 7.
> 
> 
> --
> View this message in context: http://r.789695.n4.nabble.com/Trying-to-speed-up-an-if-
> else-statement-in-simulations-tp4633725.html
> Sent from the R help mailing list archive at Nabble.com.
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.