Autoregression

An autoregressive model is a model of time-varying processes where values at subsequent time points can be regressed on previous time points. Such a regression includes an error term and potentially a constant bias term.

The noise can for example be modelled using a Gaussian, such that the mean is a regression on a previous time point with a noise according to standard deviation $\sigma$.

If each timepoint is regressed only on the previous timepoint we have Markov model. This is one particular type of autoregressive model (right!?).

Example:

$$p(x_{t+1}|x_t) = \N(x_{t+1}; \alpha x_t + \beta, \sigma^2 dt)$$

Q: What's the exact relation between Markov processes, time series models, and autoregressive models?