Identification of Linear Regression Models in the Presence of Errors in Input and Output Data
T. A. Makarova1, A. N. Tyrsin2
1 Chelyabinsk State University
2 Scientific-Engineering Center Reliability and Life of Large Systems and Machines, Ural Branch, Russian Academy of Science
toma.makarova@gmail.com
Keywords: regression models, errors in input data, Monte Carlo statistical modeling
Pages: 56-62
Abstract
The problem of constructing linear regression models in the presence of errors in input and output data is considered. A statistical test for detecting measurement errors in input data is proposed that does not require a preliminary consistent estimation of the coefficients under the assumption of the presence of errors. The test is validated by Monte-Carlo statistical simulation.
|
