

   NNoonnlliinneeaarr aauuttoorreeggrreessssiivvee mmooddeell

        nlar(Y, lags, cov=NA, method="nnreg", ...)

   AArrgguummeennttss::

          Y: The time series

       lags: A vector that specifies which lags of Y to use in
             the autoregressive function

        cov: A vector or matrix of covariates as long as the Y
             series these are additional variables that will be
             used in the regression function

     method: Name of S function to fit the nonparametric model
             e.g. nnreg tps addreg

        ...: Optional argument that as passed through to the
             regression method

   DDeessccrriippttiioonn::

        his function fits a model of the form: Y_t =  f( Y_(t-
        l1),...{},Y_(t-ld),S_t) + e_t Where e_t is assumed to
        mean zero, uncorrelated errors. Such a form is useful
        for testing whether a system is chaotic.

   VVaalluuee::

        An object of class nlar

   RReeffeerreenncceess::

        FUNFITS manual

   SSeeee AAllssoo::

        lle, predict.nlar

   EExxaammpplleess::

        # Fit the rossler series. A toy dynamical system that is chaotic
        # Use a neural network with 4 hidden units based on lags 1, 2 and 3 of
        the series.
        nlar( rossler,lags=c(1,2,3), method="nnreg",k1=4)-> out
        summary(out)
        plot( out)
        lle( out) # calculate local and global Lyapunov exponents

