Truncation Error Bounds and Convergence of Least Squares Estimates
Yoram Baram · SIAM Journal on Control and Optimization · 1980
The error resulting from truncating a data record in least squares estimation of a Gaussian process is shown to be bounded in the mean square. The bounds are shown to be easily computable for linear processes, and the truncation error is shown to be strongly diminishing under a stability condition. These results have direct implications on filtering and prediction errors, data reduction and asymptotic analysis of parameter estimates.