=== Abstract ===

In this project we study problems of inference and forecasting in autoregressive models with periodic correlation from a Bayesian perspective. Normality and unimodality assumptions are rarely verified in practice and the usual approach is to try Box-Cox transformations to obtain approximate normality and stabilize the periodic variance. More recently, mixture models were developed to take into account asymmetry and multimodality.

=== Participants ===
  - [[pessoais:ehlers|Ricardo Ehlers]]
  - Marinho Gomes de Andrade

=== Some references ===

<bibtex>
@Article{lewis,
  author = 	 {Lewis, P.A.W. and Ray, B.K.},
  title = 	 {Nonlinear Modelling of Periodic Threshold Autoregressions using TSMARS},
  journal = 	 {Journal of Time Series Analysis},
  year = 	 {2002},
  volume = 	 {23},
  number = 	 {4},
  pages = 	 {459-471}
}
@Article{shao06,
  author = 	 {Shao, Q},
  title = 	 {Mixture Periodic Autoregressive Time Series Models},
  journal = 	 {Statistics and Probability Letters},
  year = 	 {2006},
  volume = 	 {76},
  pages = 	 {609-618},
}
</bibtex>