It’s such a potent effect that Edward Tufte makes a living deconstructing it. When a great visualization is used in a presentation it has a magnetic pull on the group generating discussion and deep study (I’d count focus on a picture lasting more than 30 seconds or so as deep study in a typical meeting). In the right context, some charts can serve as standalone executive summaries.
Network models offer this kind of clear and compelling visualization, making them very well suited for the executive audience. But these models are far more than a pretty picture. They are generative models allowing for simulation and inference with efficient automated model discovery methods. Under generally accepted causality frameworks network models estimate causal effects. Any one of those attributes makes for an analytics tool worth using, but having all of them in one method is truly exceptional
To demonstrate the power of network models, this article walks through a simple network model of Intel’s financials. We consider the link between net revenue and R&D and advertising expenses as our main focus.
Let’s start with the simple network below.
With no coaching at all, I’m willing to bet you immediately understand the key points of the picture to the left: each named dot (‘node’) is a quantity, the arrows (‘edges’) indicate relationships between those quantities, and you probably guessed that the arrows point from cause to effect.
#visualization #data-science #causal-inference #data visualization
