Consistent estimation with the use of orthogonal projections for a linear regression model with errors in the variables (Q6141069)

In this paper, the author undertakes an asymptotic statistical analysis for large sample sizes, specifically focusing on consistency analysis. The strong consistency of the original total least squares solution has been proved earlier. Building upon his previous work, where he established the strong consistency of the numerical solution under row constraints, the author identifies that the case of both column and row constraints simultaneously has not been discussed in any existing study. Motivated by this, the author develops an estimator that ensures strong consistency for problems constrained in both columns and rows. Notably, the algorithm design prioritizes the properties of orthogonal projections over conventional optimization methods, leading to a natural proof of strong consistency. This approach facilitates a straightforward demonstration of strong consistency for the TLS estimator in constrained problems. It's important to note that this research does not propose a probabilistic method for efficiently solving numerical linear algebra problems. Instead, it focuses on proving strong consistency statistically using numerical techniques, thereby indicating a promising research direction. The paper also highlights the use of mathematical analysis based on orthogonal projections, which is commonly employed in perturbation analysis with condition numbers for total least squares problems. Additionally, the close connection with singular value decomposition (SVD) is noted in the statistical analysis. Specific implications for existing perturbation analysis are considered a topic for future exploration.