I have an assignment question regarding a "likelihood interval" that I don't really understand.

The question asks to consider counts of $X_i$, with $i\in \{1,...,N\}$, modelled as independent binomial variates (typo?) with constant known $n$ and mean constant unknown $p$.

Firstly I found the MLE of $p$ to be (not sure about this):$$\hat{p}=\frac{\sum_{i=1}^N x_i}{Nn}$$ We are then told to determine a $10\%$ likelihood interval based on this data where $n=10$ (why does $i$ start at $0$ now??):

$$ \begin{matrix} i & 0 & 1 & 2 & 3 & 4 & 5 \\ X_i & 2 & 0 & 1 & 5 & 3 & 3 \\ \end{matrix} $$

and also calculate a $10\%$ likelihood interval based on a single observation of $X=2$ with $n=10$.

My attempt for the first part: According to our course notes, a $c\%$ likelihood interval for $p$ is given by $$\bigg{\{} p:\frac{L(p;x)}{L(\hat{p};x)} \ge \frac{c}{100}\bigg{\}}$$

so with the the information given and taking $N=6$ (not sure due to the first count being for $i=0$ if this should be $N=5$ or not), I substituted everything into $\displaystyle \frac{L(p;x)}{L(\hat{p};x)}$ and obtained


so the $10\%$ likelihood interval becomes

$$\bigg{\{} p:(1.433678665*10^{14})*p^{14}(1-p)^{46} \ge 0.1\bigg{\}}$$

First of all, I feel like it's very unlikely that this is correct. Second of all, this isn't an interval in a form like $(x,y)$ and I don't know how to get it to be.

Does anyone have any insight? There is almost no information on the internet about likelihood intervals so I can't really make sense of this. Let me know if something is excluded or doesn't make sense and I will provide more details of my working. Also I should add that I would like to do this without the use of software if possible.


Your approach is correct. You have the right expression for the MLE.

$N$ is indeed $6$.

For the confidence interval, plot the expression $(1.43*10^{14})*p^{14}*(1-p)^{46}$ as a function of $p.$ You will see that it is concentrated around the MLE value which is $\dfrac{14}{60}=0.2333$ in this case.

From the plot you will get an idea of the confidence interval. You can solve for the exact $(a,b)$ for the confidence interval using a root finder software.

  • $\begingroup$ Thanks, just the fact that I was on the right track was really helpful. I managed to produce the plot and so on and get what seems to be the right answer. $\endgroup$ – David Pfeiffer Apr 22 at 23:24

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.