Dynamic skewness in longitudinal data models

Supervisor: Dr Miguel Juarez

Applications accepted all year round

Self-Funded PhD Students Only

About the Project

Panel (longitudinal) data enables learning the dynamics and relations of (groups of) units, strengthening the inference on both cross-sectional and dynamic parameters. The dominant approach to modelling unbounded continuous quantities of interest is to assume symmetric error terms (eg Gaussian or Student), but this assumption is often not met in practice. Several approaches have been developed to address the issue, eg generalised linear models. From a modelling perspective, it may be worthwhile considering error processes that enable interpretation of the systematic function in the model as an accessible measure of location. One option is to use skew versions of know distributions that leave the mode or median of the distribution at the origin. The main objective of this project is to explore, from a Bayesian perspective, the use of skew error distributions from well-known skewing mechanisms (eg Azzalini, 1985; Ley 2010), while allowing the parameter(s) controlling the skewness to vary over time. This implies the design of a stochastic path to controlling the process in time and exploring conditions for convergence.

Some examples include the analysis presented by Adcock et al. (2015), which considers the application of skew-elliptical distributions in the panel data context. In financial and economic time series, these distributions provide greater flexibility for modelling asymmetric data patterns. Gonzalez-Faras et al. (2004) introduced the skew-normal distribution as an approach for dealing with asymmetry in economic models, with potential applications to panel data.

Funding Notes

This project is for self-funded or externally funded students.