UKC

I'm doing some work at the moment which is essentially stressing certain variables (to represent a worsening economy) and seeing how that impacts the final outputs.

The end formula is: a*b*c = d
but a, b and c are each devised from different and fairly complicated models which each have an input value that I will stress. These inputs are, say, a1, b1 and c1, respectively.

So, for example, if the input to the model to create parameter a is a1, then I will look at increasing/decreasing a1 by certain percentages, and then assessing it's relationship with the output, d. (% change in a1 vs. % change in d).

This works fine with single variables, but if I stress different inputs together I'm unsure how best to look at these relationships. If it was just two variables, then a 3D graph would show the relationship. But stressing 3 variables leaves me a little more stuck.

There is no definite link between the inputs but they should follow some relationship. I.e. if one increases then generally the other will move up or down simultaneously. I imagine combining them in some way may work, but I'm not really sure how to do that either.

Hopefully that makes some sense. If so, any thoughts would be very welcomed.

Thanks.

In reply to James Malloch:

A 3D graph of a vs b calculated for different values of c and animated, c=time?

In reply to climbwhenready:

IMO this can only work when the audience are familiar with looking at graphs and can intuitively make sense of the time evolution and the dynamics of the underlying equations, e.g. a bunch of scientists at a conference.

I would not even dare to present data in that way to my students.

My recommendation would be pairs of illustrations showing on top selected examples of a and b (maybe as columns, a high, b low, etc.), and then below plotting d(c) for each of these examples.

CB

In reply to cb294:

Another option is to choose one variable (one with lowest range, or most accurate expected value perhaps), and treat it as a constant, taking the lowest, expected and highest value. It then reduces to three 3D graphs (kb*c = d), which you can present slowly, unlike an animation.

In reply to James Malloch:

Are these continuous variables or discrete values? If the latter is true for at least on variable then what about a matrix of x,y scatter plots?

In reply to James Malloch:
A bubble chart would do what you want.

In reply to Reach>Talent:

In 3D scatter plots the fourth dimension could also be represented by dot size or colour, both better than a dynamic representation with time as the fourth dimension.

Of course, both encodings only work for sparse, discrete data.

CB