0
$\begingroup$

I have a problem where i am uncertain how to set up a likelihood function

The number of tries before the first success follows a geometric distribution.

$$P(X=k)=(1-p)^{(k-1)}*p$$

In a sample (n=100, iid) a mean number of tries was found to be 7.

So normally without the additional information setting the max. likelihood up as normal.

But now, as i only that that the mean number of tries was seven, not the actual 100 outcomes, can i say that

$$\prod \limits_{n=1}^{100}(1-p)^6*p$$

But solving this i get a negative estimator for p, which doesn’t ring true....

It seems logical to incorporate as much information as possible into my maximum likelihood estimate....

Or could it be possible that i don’t have to integrate the information because there is a point b, (a confidence Intervall) where i would need the mean of x.

$\endgroup$
  • 1
    $\begingroup$ how do you get a negative estimate? gradient of log likelihood is $100(\frac{1}{p}-\frac{6}{1-p})$. The maximum is at the point $p=\frac{1}{7}$ having 100 identical terms just increases the sharpness of the peak. $\endgroup$ – probabilityislogic May 9 at 5:29
  • $\begingroup$ @probabilityislogic oh i made a stupid mistake, now i get p=1/7, but another question, is the Likelihood function correct? $\endgroup$ – Ang May 9 at 5:33
2
$\begingroup$

Your likelihood is correct because, if you have the entire data: $x_1,...,x_{n}$, the likelihood would be $$L(p)=\prod_{i=1}^{n} (1-p)^{x_i-1}p=(1-p)^{\sum x_i -n}p^n$$ which is your expression. Differentiating it (or log of it) will yield $p=1/7$ as your estimate.

$\endgroup$
  • $\begingroup$ Many thanks, i made a stupid mistake in my calculation! $\endgroup$ – Ang May 9 at 5:35

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.