1
$\begingroup$

I am comparing the sum of the results of a simple Poisson Distribution with the result of the Poisson Distribution Function (CDF). I thought the results were almost similar more or less. I'm using Python, but the question is clear enough regardless of using a programming language.

I want to calculate the probability having k equal to 4 or more and i'm testing the two ways. How can I get the same result with both the simple Poisson distribution and the Poisson Distribution Function (CDF)? If it is possible, I would like to place a limit of 10 on the Poisson Distribution Function (CDF), i.e. from 4 to 10

So with the simple Poisson Distribution, add from 4 to 10 (therefore imposing the maximum limit of 10):

from scipy.stats import poisson
four = poisson.pmf(4, 1.6)
five = poisson.pmf(5, 1.6)
six = poisson.pmf(6, 1.6)
seven = poisson.pmf(7, 1.6)
eight = poisson.pmf(8, 1.6)
nine = poisson.pmf(9, 1.6)
ten = poisson.pmf(10, 1.6)
a = four + five + six + seven + eight + nine + ten
print(a) 
#result is 0.078

While with the Poisson Distribution Function (CDF), I write:

b = 1-poisson.cdf(k=4, mu=1.6)
print(b)
#result is 0.023

I understand that with the simple Poisson distribution I set the limit to 10, while with the Poisson Distribution Function (CDF) there is no limit, but still the results seem too different from each other. How can I get the same result with both the simple Poisson distribution and the Poisson Distribution Function (CDF)?

The result is 0.078 with the simple Poisson Distribution, while with the Poisson Distribution Function (CDF) it is 0.023

$\endgroup$

1 Answer 1

1
$\begingroup$

In the first case you calculate $P(X\geq 4)\approx P(X=4)+P(X=5)+P(X=6)+\ldots+P(X=10)= 0.078 $

Cut off at $10$ and rounded to 2 decimal places. In the second case you calculate

$$1-P(X\leq 4)=P(X=5)+P(X=6)+\ldots= 0.023$$

Rounded to 2 decimal places.

You see that you have the extra $P(X=4)$ in the first case. In the case of discrete random variables we have the relation $P(X\geq x)=1-P(X\leq x-1)$. That means to get almost the same result like in the first case you have to calculate $1-P(X\leq 3)=0.079$ (rounded to 2 decimal places). As you have already mentioned, the discrepancy of $0.001$ comes from the cut off at $X=10$ in the first case.

$\endgroup$
8
  • $\begingroup$ I didn't understand your answer, sorry. I'm not very good at math. Basically I would like to calculate the probability of 4 or more goals. Initially I only used the simple Poisson Distribution with the limit up to 10. I didn't know about the Poisson Distribution Function (CDF). I read it casually on the web and now I'm trying to delve deeper into it. QUESTION: The logic of the Poisson Distribution Function (CDF) is the same as in my first case, but the difference is that there is no upper limit? So it's >= 4 to infinity? $\endgroup$ Commented Nov 12, 2023 at 1:35
  • $\begingroup$ @Horiatiki In the second case you have calculated $1-P(X\leq 4)$, which is $P(X\geq 5)$. But in the first case you wanted to calculate $P(X\geq 4)$. The missing terms in the first case (X=5,6,...) does not affect the result not much, only by 0,001. $\endgroup$ Commented Nov 12, 2023 at 1:46
  • $\begingroup$ So if I want to calculate >= 4, does it mean that the first case is correct, while the second case is wrong? $\endgroup$ Commented Nov 12, 2023 at 1:49
  • $\begingroup$ If yoo want to calculate $P(X\geq 4)$ you have to calculate $1-P(X\leq 3)$. So the code changes to 'b = 1-poisson.cdf(k=3, mu=1.6) ' $\endgroup$ Commented Nov 12, 2023 at 1:54
  • $\begingroup$ Great, thank you. In this way I obtain 0.07881348722971893, while with simple Poisson of the first case (From 4 to 10) I obtain 0.07881246232022453. Almost identical values. Good :) At this point a question arises: if in the first case I set the limit from 4 to 10, instead in the second case with 1-poisson.cdf(k=3, mu=1.6) what is the maximum limit? 100? $\endgroup$ Commented Nov 12, 2023 at 2:01

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .