2
$\begingroup$

I am interested in computing an upper confidence bound for the rate parameter, $\lambda$, in a Poisson process. Specifically, I have a set of observations $$ X_\text{obs} = \{(n_1,t_1), \ldots, (n_N,t_N)\} $$ where $n_k$ is the number of successes in the $k$'th trial which ran for a duration of length $t_k$.

For a given a confidence level $1 - \alpha$, I assume that what I want to do is solve for $$ \lambda_{\text{upper}} = \max \{ \lambda : P(X_\text{obs} ; \lambda) \ge \alpha \}. $$ However, this doc http://faculty.washington.edu/fscholz/Reports/poissonconfbd6.pdf suggests that I want to find something like $$ \lambda_{\text{upper}} = \max \{ \lambda : P(x \le X_\text{obs} ; \lambda) \ge \alpha \} $$ for all datasets "$x$" whose sufficient statistic is less than or equal to the sufficient statistic of the observed data.

Specifically, equation (1) in the doc sums over possible outcomes, $i = 0 \ldots k$, less than or equal to the observed number of successes, $k$. And it doesn't just compute the largest $\lambda$ for which the likelihood of the observed data is at least $\alpha$.

What is the correct way to define the upper confidence bound so that I can solve for it (either in closed form, or by numerical methods)?

Btw, I am aware that $\lambda$ is the mean number of counts per period, and so one could use deviation bounds based on the CLT to derive a UCB. However, I will be dealing with sparse counts (including zeros), so I am not sure that using deviation bounds will give me good performance. And regardless, I would like to see how to derive the UCB using the definition and/or first principles.

$\endgroup$
3
  • $\begingroup$ So for a Poisson distribution, the parameter $\lambda$ is also the mean of the distribution. This problem then resolves to estimating the confidence interval of the sample mean. $\endgroup$ Jul 12, 2019 at 19:03
  • $\begingroup$ Yes, but I'm looking for an exact computation approach, not one based on deviation bounds. This is because my counts are going to typically be very small (even zero) and the number of trials small. $\endgroup$
    – ted
    Jul 13, 2019 at 0:06
  • $\begingroup$ In this question stats.stackexchange.com/q/351320 you may find some of the logic in constructing confidence intervals. You might see the confidence intervals as being constructed as a range of values for which an α level hypothesis test would succeed So how would you construct the hypothesis test for a specific $\lambda$? $\endgroup$ Sep 12, 2019 at 18:38

2 Answers 2

1
$\begingroup$

As you notice, you can't have proper confidence bound with only the central limit theorem. In fact, confidence bounds are finite time version of the CLT.

Poisson random variables are not subgaussian, hence we don't have the classic Hoeffding bound that enjoys other subgaussian random variable (like gaussians, or bounded r.v.). However, they are sub-exponential and enjoy similar bounds. Sub-exponential have stronger tail than subgaussian (the probability of "extreme" events decay at a slower rate with the "extremeness" of the event) and this is why they need other concentration inequalities (with slower rate). This may be related to your intuition about "sparse count".

You may want to look to the detail here. To learn more about CLT and concentration result for subgaussian, you can look to chapter 5 of this book.

$\endgroup$
1
  • $\begingroup$ @Julian, the link you supplied seems to be a confidence interval for observations drawn from a Poisson, but I am trying to bound the rate parameter. Are you saying I should substitute the observed data for "X" and solve for $\lambda_\text{upper}$? $\endgroup$
    – ted
    Oct 22, 2019 at 0:35
0
$\begingroup$

Confidence intervals are frequentists tools. From a frequentist point of view, there is no such thing as confidence on the parameter. There is only one true parameter $\lambda$ which specifies your random process. This random process generates a sequence of data. Frequentists say that in other realities, the same process has generated different sequences of data. You can weight each reality with a probability. Then, you can add the weights of "realities" which have the same average. It gives you the probability of an average, i.e. the weight of realities which have this average.

The CLT tells you that the probability distribution over these averages of sequences tends to be a gaussian with a mean equals to the mean of your random process and a shriking variance as the size of the sequence grows. Confidence Intervals are tools for finite sequences. Yet, with the CLT, we understand that your $P\left(X_{\mathrm{obs}} ; \lambda\right) \geq \alpha$ does not make sense : the probability that a (almost) gaussian to take one specific value is (almost) zero. A finite probability $\alpha$ (or $1-\alpha$) can only be obtained by considering an interval of value.

Hence, CIs weight the realities which have an average which deviates by a certain quantity from the mean of your distribution. This is why we have $P\left(x \leq X_{\mathrm{obs}} ; \lambda\right) \geq \alpha$. For poisson process, the mean is your parameter $\lambda$. The average is an empirical estimate of this $\lambda$. CIs quantify how likely is the deviation between the two.

CIs involve the size of the sample $N$, the probability threshold $1 - \alpha$, and the deviation $\Delta = \hat{\lambda} - \lambda$. For a given probability $1 - \alpha$ and a sample size $N$, you can find the corresponding ""maximal"" deviation (e.g. in the link I sent you). In fine, it meant : "there is a fraction $1-\alpha$ of "realities" for which the empirical mean is at most at a $\Delta$ distance of the true parameter $\lambda$. Hence, there is no maximum on parameter (again, there is only one parameter) in confidence intervals.

I think you may be interested by Credible Intervals which is the Bayesian equivalent (there is only one reality with true data, and multiple more or less credible generating processes). The two are often very similar in shape but differs in nature/ philosophical perspective.

$\endgroup$
3
  • $\begingroup$ Hi @Julian, sorry, I am speaking loosely. By "bounding the rate parameter", I mean constructing a UCB for the true mean (which in this case corresponds to $\lambda$). People construct confidence regions for true means all the time. In fact, isn't this the mechanism behind UCB bandit algorithms? Thanks $\endgroup$
    – ted
    Oct 22, 2019 at 10:31
  • $\begingroup$ @ted : People do use quantities such as $\hat{\lambda} + c(N,\alpha)$. They are just the inversion of $\lambda - c(N,\alpha) \leq \hat{\lambda} \leq \lambda + c(N,\alpha) \implies \hat{\lambda} - c(N,\alpha) \leq \lambda \leq \hat{\lambda} + c(N,\alpha)$ which is the definition of the interval in which your average $\hat{\lambda}$ is for a fraction $1-\alpha$ of the realities. But remember that $\lambda$ is the same in all realities, the quantity which changes is $\hat{\lambda}$. $\endgroup$
    – Julien S.
    Oct 22, 2019 at 12:38
  • $\begingroup$ @ted the line of maths I have written is how you switch from 'a confidence interval for observations' to 'a UCB for the true mean'. Hence, I think that the link above is still what you need (if you are sure to think in a frequentist way). Notice that the UCB algorithm bound the frequentist regret : for a given set of arms with true means $\lambda_i$, at least $1-\alpha$ fraction of realities for which UCB1 plays these arms have a regret bounded by a certain quantity. $\endgroup$
    – Julien S.
    Oct 22, 2019 at 12:50

Your Answer

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

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