Skip to content

Evaluate output after each rep()

3 messages · Matthew MacManes, Adaikalavan Ramasamy, Bert Gunter

Matthew MacManes
# 
Hi R-Help,

I'm trying a develop a test simulation where i evaluate the probability 
of not getting a value of 100 from the function rbinom(6000, 200, .5) 
[indeed, a very small probability].  At the end of each rep, I would 
like to evaluate the output, continue with the loop if the output 
contains the value 100, stop if the output lacks a 100.

How do I get R to evaluate the output after each rep?


 >sim <- function(nn){
 >   for (ii in 1:nn){
 >      ee=rep(rbinom(6000, 200, .5), ii)
 >         if (any(ee==100))
 >      }

Thanks,
Matt MacManes
********************************************************
Matthew D. MacManes
PhD Student
UC- Berkeley
Department of Integrative Biology
Museum of Vertebrate Zoology
3101 VLSB #3140
Berkeley, CA 94720
(510)642-7782
EMAIL: macmanes at berkeley.edu
WEBSITE: http://ib.berkeley.edu/labs/lacey/
Adaikalavan Ramasamy
# 
I do not fully understand what you mean by "stop". If you mean terminate
the whole function, then something like

 sim <- function(nn, mustExist=100){

   for (ii in 1:nn){
       ee <- rep(rbinom(6000, 200, .5), ii)

       if( any(ee!=mustExist) )
           stop( paste("Iteration", ii, "did not contain", mustExist, 
                       ". Terminating function \n") )
   }
 
   ## Do something further ##
 }


But generally, I want to just resample again till I get the desired length.

 sim <- function(nn, mustExist=100){

   counter <- 0;  ii <- 1

   while( counter <= nn ){

       ee <- rep(rbinom(6000, 200, .5), ii)

       if( any(ee!=mustExist) ){
           warning(paste( "Iteration", ii, "did not contain", mustExist, 
                           ". Resampling again\n") )
       } else {
         counter <- counter + 1
         ii      <- ii + 1
         ## Do something further ##
      }
   }
 }

You can turn off the branch that returns the warnings if it gets annoying.

BTW, why do you want to use rep(.., ii) ?

Regards, Adai




If you want to count how many times out
On Mon, 2006-02-06 at 09:03 -0800, Matthew MacManes wrote:
Bert Gunter
# 
Of course I assume that everyone realizes that the simulation estimates (1 -
dbinom(100,200,p=.5))^6000. The loge of this quantity is -348, btw. So this
is trivially computable, but I assume that this is an exercise of some sort.

-- Bert Gunter
Genentech Non-Clinical Statistics
South San Francisco, CA