Let $x_1,...,x_n$ be an i.i.d. sample of geometric($p$) random variables with unknown parameter $0<p<1$. Find the likelihood of $p$, and the maximum likelihood estimate.

The pmf is $f(x;p)=(1-p)^{x-1}p$ for $x \in \{ 1,2,3,\dots \}, 0<p<1$.

How do you get from the pmf to the likelihood? I know how to go from there, I just don't know how to get it started. I need help with the first step.


The likelihood of $p$ is the probability of observing $x_1, x_2, \dots, x_n$ given that the parameter is $p$. As the trials are i.i.d., this is just the product of the individual probabilities:

$$\mathcal{L}(p) = \prod_{i=1}^n p (1-p)^{x_i - 1} = p^n (1-p)^{s_n-n}$$

where $s_k = \sum_1^k x_i$.

A maximum likelihood estimate is the maximiser of $\mathcal{L}$, which is the same as the maximiser of $\log \mathcal{L}$ (which is easier to calculate). To find this:

$$\frac{\textrm{d}}{\textrm{d}p} \log \mathcal{L}(p) = \frac{\textrm{d}}{\textrm{d}p} \left( n\log p + (s_n - n) \log (1-p) \right) = 0$$

$$\Rightarrow p = n/s_n.$$

  • $\begingroup$ Note standard notation $$ \prod_{i=1}^n p(1-p)^{x_1-1}, $$ coded as \prod_{i=1}^n, as opposed to $$ \underset{i=1}{\overset{n}{\Pi}} p(1-p)^{x_1-1}, $$ coded as \underset{i=1}{\overset{n}{\Pi}}. The former is standard and I edited accordingly. $\endgroup$ – Michael Hardy Sep 30 '17 at 14:54

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.