Create a Forum Topic by Logging In or Creating an Account

StatsPoissonCDF

Mon, 12/02/2013 - 01:14 pm

I'm having difficulty using StatsPoissonCDF, where the values don't seem quite right to me. When I generate the pdf function using StatsPoissonPDF, everything seems as it should be. For example,

StatsPoissonPDF (x = 0, lambda = 1) returns 0.368....
StatsPoissonPDF (x = 1, lambda = 1) returns 0.368....
StatsPoissonPDF (x = 2, lambda = 1) returns 0.184....

Those match theory. If I put those in a wave and integrate, all looks fine.
But, when I use StatsPoissonCDF,

StatsPoissonCDF (x = 0, lambda = 1) returns 0.736...
StatsPoissonCDF (x = 1, lambda = 1) returns 0.736...
StatsPoissonCDF (x = 2, lambda = 1) returns 0.920...

I am okay with CDF (1, 1) and CDF (2, 1), but I don't get CDF (0, 1). Shouldn't that be 0.368...? That's what the integral shows. I am getting similar behavior regardless of the choice of lambda. Did I miss something here? Is this a rounding error? I'm not sure how to get the true 0 value. I'd be grateful for a suggestion.

-Matthew

Igor

Hello Matthew,

•print statsPoissoncdf(0,1)

  0.367879

If you are getting anything different make sure you are running the latest version of IGOR.

I don't know what integral you have in mind. In the case of Poisson distribution the CDF is given by a discrete sum (see the online help).

A.G.
WaveMetrics, Inc.

December 2, 2013 at 01:52 pm - Permalink

xufriedman

Aha. I was using 6.32. 6.34 did the trick. Now the Poisson isn't acting fishy. Thanks.

-Matthew

December 4, 2013 at 01:13 pm - Permalink