[R] simulate zero-truncated Poisson distribution

[(Ted Harding)]

On 01-May-05 Peter Dalgaard wrote:
> (Ted Harding) <Ted.Harding at nessie.mcc.ac.uk> writes:
>> Something like the following code will do the job vectorially:
>> 
>>   n<-1000       # desired size of sample
>>   T<-3.5        # pre-truncation mean of Poisson
>>   U<-runif(n)   # the uniform sample
>>   t = -log(1 - U*(1 - exp(-T))) # the "first" event-times
>>   T1<-(T - t)   # the set of (T-t)
>> 
>>   X <- rpois(n,T1)+1 # the final truncated Poisson sample
>> 
>> The expected value of your truncated distribution is of course
>> related to the mean of the pre-truncated Poisson by
>> 
>>   E(X) = T/(1 - exp(-T))
> 
> There must be an easier way... Anything wrong with
> 
>  rtpois <- function(N, lambda)
>      qpois(runif(N, dpois(0, lambda), 1), lambda)
> 
>  rtpois(100,5)
> 
> ?

Well, that's neat! (When I saw that, I had to switch in
backup-brain to think clearly about qpois, but there you are ... )

Best wishes,
Ted.


--------------------------------------------------------------------
E-Mail: (Ted Harding) <Ted.Harding at nessie.mcc.ac.uk>
Fax-to-email: +44 (0)870 094 0861
Date: 01-May-05                                       Time: 21:20:43
------------------------------ XFMail ------------------------------