[R] Confidence limits for the parameter of the Poisson distribution

[Rolf Turner]

On 7/11/2008, at 9:26 AM, Hoang Trong Minh Tuan wrote:

> Hi all,
>   So far I only know one way to get the confidence limit for the  
> Poisson
> distribution is to use the look-up table given by the 2 parameter (the
> number of observation x and the confidence level, e.g. 95%) and the  
> table is
> limit by the maximum number of observations (x <= 50).
>   I know the formula to compute the CI, however, mathematically it  
> is not
> easy to do it. So, anyone know an R function to do this. Thanks
>
> Tuan.

You may find the URL

	http://www.math.mcmaster.ca/peter/s743/poissonalpha.html

useful.

	cheers,

		Rolf Turner

P.S.  The foregoing URL discusses the issues in terms of a *single*
observation.  Note that if X_1, ..., X_n are i.i.d. Poisson(lambda)
then Y = X_1 + ... + X_n is Poisson(n*lambda).

Note that the ``exact'' confidence limits are very conservative.

		R. T.

######################################################################
Attention:\ This e-mail message is privileged and confid...{{dropped:9}}