Abstract
One of the key components of causal inference relies on modeling direct and indirect effects among variables. After a review of conventional techniques including structural equation modeling (SEM) and path analysis we will discuss a hierarchical modeling alternative approach for modeling direct and indirect effects. We will also explain why the hierarchical modeling approach is preferred to SEM and Path Analysis in many regards e.g. when causal relationships among variables is largely unknown.Finally we will look at how hierarchical modeling provides a flexible tool to deal with additional sources of complexity (e.g. spatial and/or temporal dependence structure in the data) as it allows for exploring relationships among variables using conditional modeling as opposed to joint modeling.

Presenters
Ali Arab