Multifractional, multistable, and other processes with Prescribed local form (Q1028614)

In this interesting paper a general method for constructing stochastic processes with prescribed local form is presented. It is applied to certain Poisson sums to construct processes where both parameters, local regularity (\(h\)) and jump intensity (\(\alpha\)), vary in time in a prescribed way. The authors' construction allows known localisable processes \(\{X(t,v) : t \in \mathbb{R} \}\) to be pieced together over a range of \(v\) to yield a localisable diagonal process \(Y=\{X(t,t) : t \in \mathbb{R}\}\) with local form depending on \(t\). The interplay of \(X(.,v)\) for \(v\) in a neighbourhood of \(u\) is crucial to the local behaviour of \(Y\) near \(u\) and criteria that guarantee the transference of localisability from the \(X(.,v)\) to \(Y\) are derived. The authors illustrate their method with multifractional processes, multistable processes and multifractional multistable processes where both the local self-similarity index (\(h\)) and the stability index (\(\alpha\)) vary.