In statistics and signal processing, an autoregressive (AR) model is a representation of a type of random process; it describes the relationships of values at prior times to those of current times (as well as for covariates) during time-varying processes in nature, economics, etc.