Please use this identifier to cite or link to this item:
|Title||Non-Gaussian maximum entropy processes|
The maximum entropy principle provides one of the bases for specification of complete models from partial information. It was introduced to time series by the influential work of Burg in the late 60’s and early 70’s (Burg, 1975). The principle postulates that among all processes consistent with the prior information one with the highest entropy rate should be chosen. The prior information usually consists of the values of the autocovariance function (acvf) for some lags or, more generally, pairs of times (t, s)∈ I where I is a subset of N2. The aim is to find a model with maximum entropy rate whose autocovariance function has the given values on I. The full problem thus consists of specifying values of the autocovariance function for all pairs (t, s)∈ I (ie completing or extending it) and a probabilistic structure such that the entropy rate is maximal. In second order estimation the distribution part is often ignored.
|Item link||Item Link|
|Files in this item|
|Iqelan, Bisher M._4.pdf||178.7Kb|