1: \begin{abstract}

2: In the matter of selection of sample time points for the estimation

3: of the power spectral density of a continuous time stationary

4: stochastic process, irregular sampling schemes such as Poisson

5: sampling are often preferred over regular (uniform) sampling. A

6: major reason for this preference is the well-known problem of

7: inconsistency of estimators based on regular sampling, when the

8: underlying power spectral density is not bandlimited. It is argued

9: in this paper that, in consideration of a large sample property like

10: consistency, it is natural to allow the sampling rate to go to

11: infinity as the sample size goes to infinity. Through appropriate

12: asymptotic calculations under this scenario, it is shown that the

13: smoothed periodogram based on regularly spaced data is a consistent

14: estimator of the spectral density, even when the latter is not

15: band-limited. It transpires that, under similar assumptions, the

16: estimators based on uniformly sampled and Poisson-sampled data have

17: about the same rate of convergence. Apart from providing this

18: reassuring message, the paper also gives a guideline for

19: practitioners regarding appropriate choice of the sampling rate.

20: Theoretical calculations for large samples and Monte-Carlo

21: simulations for small samples indicate that the smoothed periodogram

22: based on uniformly sampled data have less variance and more bias

23: than its counterpart based on Poisson sampled data.

24: \end{abstract}

25: