I have $n$ realisations $s_1,\, \dots , s_n$ of random variables $S_1,\, \dots, S_n$ which are assumed to be i.i.d. with unknown distribution. These measure the time between events.

I want to calculate the probability that the data can be modelled by a renewal process: $(X_t)_{t\geq0}$ where $X_t$ is the total number of jumps by time $t$.

In particular, I am using the elementary renewal equation $\lim_{t\rightarrow \infty} \frac{E[X_t]}{t} = \frac{1}{E[S_i]}$ and I want the probability that the observed values of the LHS and RHS agree for a given $(n, t)$.

Do I want to some how construct a confidence interval for the statistic $\frac{E[X_t]}{t}$ and test whether $\frac{1}{E[S_i]}$ lies in this interval?

I guess what I'm asking, is how can I calculate the error in the strong law of large numbers for a given $n$, I think?

Apologies for the confusing description, my statistics isn't the best!

  • 1
    $\begingroup$ $E[X_t]/t$ is not a statistic. $X_t/t$ might be one, but its expectation isn't. $\endgroup$ – Glen_b Apr 2 '13 at 9:06
  • $\begingroup$ If $S_1, ... S_n$ are (assumed to be) i.i.d., then $(X_t)$ is a renewal process. No ? $\endgroup$ – Stéphane Laurent Apr 2 '13 at 9:51
  • $\begingroup$ Ah, yes. So what I am really testing is whether $S_1, \dots, S_n$ are i.i.d. I guess? $\endgroup$ – Dave White Apr 2 '13 at 10:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.