The number of accidents in a week follows a poisson distribution with mean $\lambda$.

Likelyhood is given as $$L(\lambda)=\frac{ \lambda^{\sum_1^n x_i } e^{-n\lambda}} { \prod x_i!}$$

However only a proportion p, are reported and each accident is reported with probability p, independent of all others.

How would you modify the likelyhood function to take account of this?


The probability without the report issue is $$ P(X = n) = \frac{\lambda^n e^{-\lambda}}{n!} $$

including the report issue it becomes for $n\neq 0$: $$ P(X = n) = \sum_{m=n}^\infty \frac{\lambda^m e^{-\lambda}}{m!} p^n(1-p)^{m-n} $$ (this is the probability of $m$ accidents, $n$ of which being reported).

Then just write $$L(\lambda) = \prod_{k=1}^n \sum_{m=n}^\infty \frac{\lambda^m e^{-\lambda}}{m!} p^n(1-p)^{m-n}$$

I'm not sure you can get any better expression.

  • $\begingroup$ So for n, the number of reported accidents, do they still follow a poisson distribution? I can't really tell just by looking at the expression. $\endgroup$ – Dylan Morgan Oct 20 '14 at 22:45
  • $\begingroup$ I don't think the expression does have the right shape: it has the form $[p/(1-p)]^n F(n)$ where F(n) is the rest of an exponential series. But this is not a proof... $\endgroup$ – mookid Oct 20 '14 at 22:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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