Model Selection in Mixture Regression Analysis--A Monte Carlo Simulation Study

Sarstedt, Marko und Schwaiger, Manfred (2008): Model Selection in Mixture Regression Analysis--A Monte Carlo Simulation Study. 31st Annual Conference of the German Classification Society, Albert-Ludwigs-Universität Freiburg,, 7.-9. März 2007. Preisach, Christine; Burkhardt, Hans; Schmidt-Thieme, Lars und Decker, Reinhold (Hrsg.): In: Data Analysis, Machine Learning and Applications, Berlin; Heidelberg: Springer. S. 61-68

Abstract

Mixture regression models have increasingly received attention from both marketing theory and practice, but the question of selecting the correct number of segments is still without a satisfactory answer. Various authors have considered this problem, but as most of available studies appeared in statistics literature, they aim to exemplify the effectiveness of new proposed measures, instead of revealing the performance of measures commonly available in statistical packages. The study investigates how well commonly used information criteria perform in mixture regression of normal data, with alternating sample sizes. In order to account for different levels of heterogeneity, this factor was analyzed for different mixture proportions. As existing studies only evaluate the criteria's relative performance, the resulting success rates were compared with an outside criterion, so called chance models. The findings prove helpful for specific constellations.

• Dokumententyp: Konferenzbeitrag (Paper)
• Fakultät: Betriebswirtschaft > Institut für Marktorientierte Unternehmensführung
• Themengebiete: 300 Sozialwissenschaften > 330 Wirtschaft
• Ort: Berlin; Heidelberg
• Sprache: Englisch