## Problem Statement

Let thus , we construct a stochastic process as , this kind of random variable are called hypoexponential random variables. We define a counting process such that . This allows to model a certain relation between events, for example to have a shorter expected waiting time for a second event if the first event was observed.

**Statement 1:** is no Poisson process.

To see this let and then while . Since the density of is given by with the process is non stationary and can thus not be a Poisson process.

**Statement 2:** The (hypoexponential) distribution of is given by for and else where and thus .

(1)

## Estimation

To estimate we construct a different process. For fixed , let be sorted descending by size. this is the amount of all processes where the -th event was reached at time . Let , here is determined by the ordering of , this is also often referred to as exposure. is now estimated with .

The sum of order statistics exponential random variables is given by [1] thus the joined distribution is given by

(2)

Note that and are not independent, in particular

(3)

(see Statement 2), if the realisation of is relatively small than for given the probability that the event was observed is higher. For simplicity, in the following, we will consider a special case where , then and the sum of exponential random variables is given by , thus

(4)

Therefore the estimator is unbiased while the MSE is given by . Which reflects the well-known dichotomy between unbiasedness and minimal MSE.

For the general case

(5)

thus we face a slight bias, the cause of which is when there are too few observations available.

[Bibtex]

```
@Inbook{Nagaraja2006,
author="Nagaraja, H. N.",
editor="Balakrishnan, N.
and Sarabia, Jos{\'e} Mar{\'i}a
and Castillo, Enrique",
title="Order Statistics from Independent Exponential Random Variables and the Sum of the Top Order Statistics",
bookTitle="Advances in Distribution Theory, Order Statistics, and Inference",
year="2006",
publisher="Birkh{\"a}user Boston",
address="Boston, MA",
pages="173--185",
abstract="Let X(1)<...
```

` `