

seasest(timeslab)                            R Documentation

_C_a_l_c_u_l_a_t_e _B_o_x_-_J_e_n_k_i_n_s _E_s_t_i_m_a_t_e_s _f_o_r _a _S_e_a_s_o_n_a_l _A_R_I_M_A _M_o_d_e_l

_D_e_s_c_r_i_p_t_i_o_n_:

     Calculate Box-Jenkins Estimates for a Seasonal ARIMA
     Model

_U_s_a_g_e_:

     seasest(y,ords,coeffs,lags,back,maxit=50,eps=0.000001)

_A_r_g_u_m_e_n_t_s_:

       y: Array of length $n$ containing the data.

    ords: An array of length 5 containing the full and sub-
          set AR orders, followed by the full and subset MA
          orders, followed by a 1 if a constant term is in
          the model or a 0 if it is not.

  coeffs: An array containing starting values for the coef-
          ficients that are included in the model in the
          order full AR, subset AR, full MA, subset MA and
          the mean of {`y'}.

    lags: An array containing the lags (if any) in the
          model.  If both the subset AR and MA orders are
          zero, no array called {`lags'} need be formed, but
          an argument must be included.

    back: An integer containing the number of back forecasts
          to used in determining initial values in the
          recursion used in evaluating the sum of squares of
          residuals functions $(>=0)$.

   maxit: An integer containing the number of iterations to
          allow in the estimation procedure.  If {`maxit'}
          is negative, then -{`maxit'} iterations are
          allowed and the values of the coefficients for the
          successive iterations are displayed on the screen.
          If {`maxit'} is 1 then {`SEASEST'} only evaluates
          {`rvar'} and {sl{sds}}.

     eps: Real scalar containing a convergence criterion.
          If the maximum value of successive iterates dif-
          fers by less than {`eps'}, then {`SEASEST'} judges
          that the algorithm has converged.

_V_a_l_u_e_:

 seasest: returns a list containing the following five ele-
          ments:

  coeffs: Array containing the final values reached for the
          parameters in the iterative process.  {`coeffs'}
          is not changed from input if {`maxit'}$=1$.

       e: Array of length $n$ containing the one step ahead
          prediction errors corresponding to the $n$ values
          of {`x'}.

     ier: An integer variable indicating whether or not con-
          vergence was achieved (0 means yes, 1 means no),
          if a singular matrix was encountered (2), or
          whether the algorithm could not continue even
          though convergence was no reached (3 or 4).  If
          this final alternative happens, different starting
          values or convergence criteria may lead to conver-
          gence.

      rv: Real scalar containing an estimate of the error
          variance.

      se: An array containing the standard errors of the
          estimates.

