Weekend update: In quest of 3d graphic for risk perception distributions

In response to the scatter plots from “politicization of science Q&A” post, @thompn4 on twitter (optimal venue for in depth scholarly exchange) observed that it would be nice to have a three-dimensional graphic that combined partisanship, risk perception, and science comprehension (or perhaps two risk perceptions — like nuclear and global warming — along with science comprehension or partisanship) into one figure.

Great idea!

I supplied @thompn4 with data, and he came up with some interesting topographical plots.

Pretty cool!

But these are all 2 dimensional — and so fail to achieve what I understand to be his original goal– to have 3d representations of the raw data so that all the relevant comparisons could be in one figure and so there’d be no need to aggregate & split the data along one dimension  (as the science comprehension plots do).

When I pressed him, he came up with a 3d version, but with only 2 dimensions of individual difference — science comprehension & risk perception:

Really great, but I want what he asked for — three graphic dimensions for three dimensions of individual difference.

I’ve been fumbling with 3d scatter plots.  Here’s ideology (x), risk perception (y), and science compression (z)– with observations color-coded, as in 2d scatter plots, to denote perceived risk of global warming (blue = low to red = high):

Not great, but it gets at least a bit better when one rotates the axes counter-clockwise:

I suspect a topographical or wireframe will work better than a scatter plot — but that’s something beyond my present graphic capabilities.

In the end, too, the criteria for judging these 3d graphs, in my view, is whether they enable a curious, reflective person readily to discern the relevant information — and in particular the existence of an important contrast.  Being ornate & attention-grabbing are not really the point, in my view. So far not clear to me that anything really improves upon the original 2 graphic solution.

If anyone else wants to try, feel free.  The data are here. Please do share your results — you can email them to me or post them somewhere w/ URL I can link to.


1. The data are tab delimited.

2. Zconservrepub is a standardized sum of 7-point partyid & 5-point liberal-conservative ideology, valenced toward conservative/republican.

3. scicomp_i is score on a science-comprehension assessment (scored with item response theory; details here)

4.GWRISk & NUKERISK are “industrial strength risk perception measures” for “global warming” & “nuclear power. Each item is 0-7: 0 “no risk at all”; 1 “Very low risk”;  2 “Low risk”; 3 “Between low and moderate risk”; 4 “Moderate risk”; 5 “Between moderate and high risk”; 6 “High risk”; 7 “Very high risk”

There are 2000 observations total.  Some observations have missing data.

Leave a Comment