No label defined (Q4356557)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 1069596
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 1069596 |
Statements
28 January 1998
0 references
ExpAR models
0 references
exponential family
0 references
stationary homogeneous diffusion process
0 references
ARCH models
0 references
nonlinear state space model
0 references
time series
0 references
0.8047894835472107
0 references
It has been shown that a simple first lag order ExpAR model can generate time series with various types of marginal distributions by varying the two model coefficients. A close relationship between ExpAR models and the exponential family of distributions is pointed out. It is shown that for a given distribution of the exponential family we can have a special type of diffusion process whose diffusion coefficient is constant and whose marginal distribution is equivalent to the given distribution. Any such stationary homogeneous diffusion process with constant diffusion coefficient can be closely approximated by a discrete-time nonlinear autoregressive model whose autoregressive coefficient is a smooth function of the process and approaches a constant for \(|x_t|\to\infty\). This means that for a given distribution of the exponential family we can introduce an ExpAR model whose marginal distribution is very close to the given distribution. For the analysis of time series with a stable marginal distribution, heteroscedastic nonlinear time series models such as ARCH models are used. It has been shown that an ExpAR model combined with an instantaneous variable transformation is useful for the modelling of such heteroscedastic time series. A nonlinear state space model was introduced as an extension of the ExpAR model. The model-set defined by the nonlinear state space representation includes \(\text{ARMA} (p,p-1)\) models as a subset. Thus it can be regarded as a natural nonlinear extension of linear Markovian models for time series. For the entire collection see [Zbl 0846.00010].
0 references