1. Inference. Estimates
stationary p.p. {N(t)}, rate pN , observed for 0<t<T
First-order.

2. Asymptotically normal.

4. Theorem. Suppose cumulant spectra bounded, then N(T) is asymptotically N(TpN , 2Tf2 (0)).
Proof.
The normal is determined by its moments

6. Second-order.

9. Bivariate p.p.

10. Volkonski and Rozanov (1959); If NT(I), T=1,2,… sequence of point processes with pNT  0 as T   then, under further regularity conditions, sequence with rescaled time, NT(I/pNT ), T=1,2,…tends to a Poisson process.
Perhaps INMT(u) approximately Poisson, rate TpNMT(u)
Take:  = L/T, L fixed
NT(t) spike if M spike in (t,t+dt] and N spike in (t+u,t+u+L/T]
rate ~ pNM(u) /T  0 as T  
NT(IT) approx Poisson
INMT(u) ~ N T(IT) approx Poisson, mean TpNM(u)

11. Variance stabilizing transfor for Poisson: square root

12. For large mean the Poisson is approx normal

13. Nonstationary case. pN(t)

14. A Poisson case.
Rate
(t) = exp{ +  cos(t + )}
 = (,,,)
log-likelihood
l() =  log ( i) -  (t) dt < t < T