1 Why interactive sites?

We often find ourselves looking at data and realize that there are a million and one ways to visualize it. As scientists, we are supposed to decide which story to pluck out to make our argument. What this means is that we end up settling on 1-2 plots, each of which has only 2-4 dimensions (e.g. x and y axis, and sometimes a coloring or faceting that adds richness to those dimensions).

While part of being a scientist is deciding on what dimensions are important for the story you think your data tells, part of being a science communicator, teacher, and an open data advocate is finding ways to share the multi-dimensionality of a story. Interactive data sites can help us do that, particularly for communicating to wider audiences and teaching.

1.1 Communication and personalization

Often data doesn’t have just one story. Readers don’t want to know about the average, or how something compares against a baseline that is irrelevant for them. People want to know how data relates to their lives and their interests. Interactive data sites let users be in control of what they want to see. For example:

• UC Davis TA/GSR pay estimator lets students see their potential wages based on their own starting conditions
• State of Organic Seed Report interactive data lets stakeholders view statistics most relevant to them (e.g. crops, regions, etc.)

1.2 Teaching

Interaction is also an incredible tool for teaching. It allows us to see outcomes across several scenarios, see how relationshps shift when new variables are added, etc. Some examples I like include:

• Michael Levy’s demonstration of how network parameters and hyperparameters relate to the shape of a network in a certain modeling technique
  
• Chelsae Pelleriti’s demonstration of how different classification algorithms create different decision boundaries and vary based on parameters¹

---

1. P.S. Chelsea has one of the most entertaining online presences re: stats and coding @ChelseaParlett↩︎