Controlled Data: Generation, Testing, Modeling and Impact Gauging

Abstract

Several test procedures are available to check the admissibility of underlying assumptions of statistical procedures. Such procedures have potential to confirm or deny the adequacy of the assumption. In real data sets we do have indications and symptoms regarding assumptions but we are not sure about their presence and/or intensity. In the circumstances we have to rely on whatever is being told by the test procedures. To check the admissibility and potentials of these test procedures we need to apply them on the data which have been cloned to possess certain characteristics at known level. This article concentrates on the generation of controlled data with known level of multicollinearity, heteroscedasticity, and autocorrelation and establishing relation of common test procedures with the intensity of the violation of basic assumptions of linear models.