The chapter considers the estimation algorithm of AR parameters in the presence of white observation noise. The estimation algorithm is iterative because the estimation of noise variance and parameters is sequentially repeated until some convergence criterion is satisfied. The algorithm is built upon the bias correction principle (Stoica et al. 1987; Zheng 1999; Zheng 2000). The idea is to use the HO(Y-W) equations (i.e., to use more autocorrelation samples), sequentially estimate the noise variance, and to use these variance estimates for the bias correction.