Analysis and lattice implementation of Extended Instrumental Variable (EIV) methods for high resolution spectral analysis are presented. The performance of EIV is improved by using prefilters and the unbiasness of EIV is proved by using the fact that residual processes are white. We derive the order and time update formulas for the covariance lattice algorithm which is particularly useful in case of short data or nonstationary processes. The ARMA model can be modeled as two channel AR processes. Using this model, the lattice algorithms of EIV are derived. Computer simulations are performed to show the usefulness of the proposed algorithms.