Characterization of Discrete Time Scale Invariant Markov Processes S. Rezakhah and M. Modarresi Abstract Scale invariant processes have recently drawn attention of many researchers. Con- sidering discrete samples requires new tools, which we have provided as some spe- cial quasi Lamperti transform. We study a discrete scale invariant Markov process {X(t),t R + } with scale l> 1 and consider to have some fix number of observa- tions in every scale, say T , and to get our samples at discrete points α k ,k Z, where α is obtained by the equality l = α T . So we provide a discrete time scale invariant Markov (DT-SIM) process X(·) with parameter space {α k ,k Z}. We show that the covariance structure of DT-SIM process is characterized by the values of {R H j (1),R H j (0),j =0, 1,...,T 1}, where R H j (k) is the covariance function of j th and (j + k)th observation of the process. We introduce discrete scale invariant au- toregressive process of order p, DSIAR(p) with time varying coefficients. We present two examples of DT-SIM process as DSIAR(1) and discrete time simple Brownian motion and justify our characterization result. In correspondence to our DT-SIM process with scale α T , we introduce T -dimensional self-similar Markov process. AMS 2000 Subject Classification: 60G18, 60J05, 60G12. Keywords: Discrete scale invariance; Wide sense Markov; Multi-dimensional self- similar processes. 1 Introduction 2.1The notion[10] of scale invariance or self-similarity is used as a fundamental property to handle and interpret many natural phenomena, like textures in geophysics, turbulence of fluids, data of network traffic and image processing, etc [1]. The idea is that a function is scale invariant if it is identical to any of its rescaled functions, up to some suitable renormalization of its amplitude. Faculty of Mathematics and Computer Science, Amirkabir University of Technology, 424 Hafez Av- enue, Tehran 15914, Iran. E-mail: rezakhah@aut.ac.ir (S. Rezakhah), namomath@aut.ac.ir (N. Modarresi). 1