Performance contours of autoregressive estimates

Given a pth-order AR (autoregressive) process, the authors obtain an explicit formula that relates the true parameters and their estimates (or distorted versions) to the variance of the prediction error process. For a second-order AR model they determine contours within which the parameter estimates must lie, corresponding to a set of prediction error variance values. Conversely, for a given set of parameter estimates, they obtain contours within which the true parameters would lie, corresponding to a set of prediction error variance values. These results provide some interesting insights about AR modeling