The central aim of the proposed project is to develop a new framework for estimating measurement error which can be applied to longitudinal data structures. Using an integration of the classic factorial experimental design and latent variable modelling, the new approach will enable estimation of not only random error but also a number of systematic errors such as: the method effect, social desirability and acquiescence in longitudinal data. In addition to the inclusion of multiple types of errors within a single framework it will also be possible to estimate how they change over time. This will enable powerful and flexible error corrections in longitudinal datasets. The project also, therefore, seeks to enhance and add value to the growing number of longitudinal datasets, both in the UK and internationally, which are increasingly seen as essential to addressing key policy research questions.
The programme of work comprises three inter-related objectives, to:
- develop a new method for detecting and correcting measurement error in longitudinal data
- pply the new method to investigate the impact of measurement error on the analysis of attitudes towards immigrants in longitudinal and cross-national data
- encourage take up of the new method by users of longitudinal datasets through dissemination and engagement activities
These objectives will be delivered through the analysis of the 'attitudes towards immigrants' scale which have been administered in three consecutive waves of the Understanding Society Innovation Panel and the German Socio-Economic Panel Innovation Sample. These data contain the experimental design which is required to estimate the measurement models proposed here.