Advanced topics in sensitivity analysis
Sensitivity analysis in causal inference has traditionally zeroed in on binary treatment effects within straightforward models. Yet, as we march forward in time, there’s an emerging consensus on the necessity to push these methods into more intricate arenas. This includes dealing with continuous treatments, interventions that span multiple levels, and even weaving in machine learning (ML) techniques. Let’s dive into these advanced areas.
Venturing beyond binary treatment
First, we have continuous and multi-level treatments. The classic picture of sensitivity analysis often paints treatments in stark black and white - you either get the treatment or you don’t. But the real world isn’t always so binary. Treatments can vary in dosage or come in different forms (think varying educational programs). To tackle these realities, sensitivity analysis is stretching its muscles to understand how changes in the intensity...
