On the Bartlett spectrum of randomized Hawkes processes (Q2869104)

Hawkens introduced a class of point processes which he called self-exciting processes (SEP). Nowadays, they are called linear SEP. Generalized Hawkes processes with a nonlinear rate function are studied in [\textit{P. Brémaud} and \textit{L. Massoulié}, J. Appl. Probab. 38, No. 1, 122--135 (2001; Zbl 0983.60048)] and [\textit{P. Brémaud} et al., J. Appl. Probab. 39, No. 1, 123--136 (2002; Zbl 1005.60062)]. In the present paper, randomized Hawkes processes are introduced. The authors study their Bartlett spectra with the goal to provide an example of a point process with singular Bartlett spectrum or long-range dependence. Long-range dependence or long-memory phenomena in stochastic geometry is a rapidly developing subject in probability and statistics, see [\textit{P. Doukhan} (ed.) et al., Theory and applications of long-range dependence. Boston, MA: Birkhäuser (2003; Zbl 1005.00017)]. This book contains an outstanding survey of the field, in particular, it discusses different definitions of long-range dependence of stationary processes in terms of the autocorrelation function (the integral of the correlation function diverges) or the spectrum (the spectral density has a singularity at zero). In the point processes context, the definition of long-range dependence has to be reconsidered by transforming the spectrum into the Bartlett spectrum. The present paper presents a new version of point processes with the singular Bartlett spectrum originated from SEP processes by randomizing a parameter of the exciting function. The authors show similarities and differences of the singular property of the Bartlett spectrum with the one of the Bochner-Khintchine spectrum (as indication of long-range dependence) for stationary processes used in different models with randomization of parameters before, see [Brémaud and Massoulié, loc. cit.].