Statistical Modeling, Causal Inference, and Social Science: Surface vs. Contour Plots for the Presentation of Three-Dimensional Data

Surface vs. Contour Plots for the Presentation of Three-Dimensional Data

Chris Zorn sends along this. A rare paper about graphics that offers data as well as opinion!

Sergei Dorn

Except that the contour and the surface for the linear case don't match (the contour plot seems to be upside-down)... Which leads to the very strange results obtained in this case with the students subjected to the surface outperforming those having the contour.

In this experiment, it's obviously easier to read the values on the contour plot. I am wondering what the results would have been with an improved display for the surface plots (colors, Y-scale directly on the surface, etc.) .

Hadley

I would have thought that the obvious comparison would be between contour plots and image plots (i.e. where z-position is mapped to colour).

Testing the ability of students to accurately decode values on the graphic is a good start, but that's not really the purpose of a graphic. It would be interesting to see a similar experiment performed on more gestalt-based questions - i.e. how many peaks are there, is the surface smooth or rough etc.

C. Zorn

Sergei:

Wow.

You're right.

*smacks forehead*

I'd better fix that...

Thank you!

Bill Harris

There's always "true" 3-D: see slide 20 of http://facilitatedsystems.com/weblog/sdforcheapskates.beamer.pdf (simple to do in J).

Phil

I give up: how do I see the paper? I can read the abstract (as 332 other viewers have apparently done) but clicking "Go to Download Document" gets me no joy, and the icon says the paper has been downloaded "0 times" so I guess I'm not the only one with a problem. Yet, Sergei was apparently able to get it. I guess I have failed a simple IQ test. Sigh. Well, it's not the first time.

C. Zorn

It's not available because...well...I pulled it. I'll repost once I fix the mistake that Sergei so helpfully pointed out.