This project is concerned with a methodological question about how multilevel models are used in applied social research. Multilevel models are important statistical methods that are often used in social science projects. Nevertheless, when multilevel models are applied, they frequently violate a statistical assumption about their specification, namely that the estimated random effects are uncorrelated with explanatory variables (‘NCRX’). If there is a correlation, it has been demonstrated that parameter estimates could be biased and/or inefficient. Although the statistical issues surrounding the NCRX assumption have been demonstrated, there remain many research applications where the assumption is not fully explored, and there are divergences between social science disciplines in how seriously the assumption is taken.
This PhD project will (I) review methodological and applied research literature on the NCRX assumption across social science disciplines, (2) analyze simulated data to explore the importance of the assumption in different contexts, and (3) use the ‘Understanding Society’ secondary social survey (special license versions) to carry out two case studies on socio-economic inequalities by occupations and health inequalities by localities in which the NCRX assumption is likely to be relevant. These two case studies are tailored to complement the applied research interests of the supervisors.