1
$\begingroup$

I have a p-value definition as follows and I would like to implement it.

enter image description here

The values of n1, n2, n3, n4, n5 and N are as follows.

n1 = 102
n2 = 95

Calculating λ:

n3 = 1350
n4 = 3588
n5 = 886
N = 10382

So my λ would be 421.41

I followed scipy.stats.poisson documentation: https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.poisson.html

from scipy.stats import poisson

However, scipy provides number of methods to calculate poisson value as;

rvs(mu, loc=0, size=1, random_state=None)   Random variates.
pmf(k, mu, loc=0)   Probability mass function.
etc.

I am still not sure what method I should use to get the summation of poisson vales from i = n1+n2, ...., ∞.

NOTE: I am not limited to scipy and would like to receive answers using python or R.

I am happy to provide more details if needed.

$\endgroup$

1 Answer 1

1
$\begingroup$

The p-value definition that you want to implement is the probability that a Poisson random variable with parameter $\lambda$ is greater than $n_{1} + n_{2}$. If $X\sim Poisson(\lambda)$, then $\sum_{i=n_{1} + n_{2}}^{\infty} e^{-\lambda} \frac{\lambda^{i}}{i!}=\mathbb{P}(X \geq n_{1} + n_{2}) = 1 - \mathbb{P}(X \leq n_{1} + n_{2} - 1)$.

In R, you can use the ppois(q, lambda, lower.tail) function with $q = n_{1} + n_{2}$, $lambda = \lambda$ and lower.tail = FALSE to calculate $\mathbb{P}(X \geq n_{1} + n_{2})$.

I have never used Python, but the page you link to has two function that I think that you can use. You can use the cdf(k, mu, loc=0) function with $k = n_{1} + n_{2}-1$ and $mu = \lambda$. Alternatively, you can use the sf(k, mu, loc=0) function with $k = n_{1} + n_{2}$ and $mu = \lambda$.

$\endgroup$
1
  • $\begingroup$ Thanks a lot. I really appreciate your recommendations in python :) $\endgroup$
    – EmJ
    Commented May 14, 2019 at 12:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.