Title: Behavior of rank tests and estimates in measurement error models
Authors: Radim Navratil - Masaryk University Brno (Czech Republic) [presenting]
Abstract: Measurement error models (also called errors-in-variables models) are regression models that account for measurement errors in the independent variables (regressors). These models occur very commonly in practical data analysis, where some variables cannot be observed exactly, usually due to instrument or sampling error. Sometimes ignoring measurement errors may lead to correct conclusions, however in some situations it may have dramatic consequences. Behavior of standard rank procedures (both tests and estimates) in measurement error models will be investigated. The main goal is to investigate if classical rank tests and estimates stay valid and applicable when there are some measurement errors present and if not how to modify these procedures to be still able to do some statistical inference. Finally, performance of the tests and estimates will be illustrated on practical examples and simulations.