Hey, there have been talks about using D3. Would like to start a thread on different visualization techniques, for different application domains, e.g. Token Engineering would like to “see stakeholder grpahs” - I think this can be achieved by wrapping networkX accordingly. Any ideas, pointers?
2 Likes
One pointer that I had recently, is that the visualization pattern domains depends strongly on the data structure itself, and as such, there is a non-trivial structure description phase that can save a lot of time and work in regards to seeing what plots make sense.
To borrow some concepts, all data can be roughly defined as being one of those two things: category or measurement. Category are ‘qualities’ or ‘properties’, while measurements are ‘quantities’ or ‘variables’.
In a simulation context, the sweeped parameters would normally be categories, while the state variables would be measurements, although this is not a rule.
Having this is mind, it is a general knowledge on the viz community that one consistent way of fitting data into visualization patterns generally is by having a ‘measurement table’ which fits in this category vs measurement paradigm, and to choose the dimensions, groupers and filters accordingly. Tableau uses this kind of procedure a lot on their viz philosophy.
One problem when doing complex models, is that you can have data with high-dimensionality which don’t fit directly on this ‘measurement table’ scheme. There is no trivial way to handle that, and I suppose that it’s kind of the reflex of what happens when you try to project 5D into 2D.
This is some random thoughts about the subject. I’ll come here again when I have some visual examples
2 Likes