MLE of exponential mean from summary totals

Question

Latencies for a network service are modeled as independent exponential variables with unknown mean $\theta$ , density $f(x \mid \theta) = \frac{1}{\theta} e^{-x/\theta}$ for $x > 0$ . You are told only that $n = 20$ latencies were recorded and that their sum was 150 milliseconds; the individual values were not kept.

What is the maximum likelihood estimate of $\theta$ ?

Accepted Answer

The MLE of an exponential mean depends on the data only through the total, because the log-likelihood is \ell(	heta) = -n\ln	heta - \frac{1}{	heta}\sumi xi, and solving \ell'(	heta) = 0 gives \hat	heta = \frac{1}{n}\sumi xi. The sum \sumi xi is a sufficient statistic, so knowing the total and the count is all you need: \hat	heta = \frac{150}{20} = 7.5 	ext{ milliseconds}. Takeaway. The exponential MLE uses only the sum and the count, so summary totals are enough to compute it.

MLE of exponential mean from summary totals

Your answer

Discussion

Learn the concepts

Related questions