Excel does not make native 3D XYZ charts. Andy Pope (http://andypope.info) has developed a system that uses a ton of trig to convert X-Y-Z coordinates to a 2D medium with the appearance of 3 dimensions. Sliders let the user change perspective. It’s a good hack

I have seen effective 3D scatter charts; they are dynamic and interactive. I’ve never seen anything meaningful done in Excel.

In addition to the factors mentioned (new ballparks, stadium size, etc.), I also immediately thought of:
(1) What impact does the opponent have on attendance? I imagine this week’s games in San Francisco with the Rockies, teams that are neck-and-neck for a playoff position, is much different than the Giants normally see. I also know, regardless of how my Rockies are performing, if someone like the Yankees are in town, attendance is through the roof.

(2) What impact does the team’s position in the playoff race have and would there be a way to see that over time. Again, with my Rockies, I’ll be their attendance in the second half of the season when they started making a run for the playoffs is much better than when they looked dismal earlier in the year.

Just some thoughts.

Interesting post, good stuff.

bit of a spurious aside, but I would be interested to hear your views on ‘3d scatter plots’. do you think they have value?
Excel is perhaps not the best environment for one, but assuming you could quickly output a 3d model, the camera angle of which could be manipulated by the user, would you opt to give them full control to see the data from any angle, or would you rather restrict them to 3 views of the ‘front’, ‘top’ and ’side’ of the ‘box’ (ie. 3 seperate xy charts).

would you chose another method entirely to display 3 things which do not share a common scale (unit).

-Matt

I didn’t bother with the residuals because there are many other factors to take into account, and they would affect the residuals in ways I was not prepared to investigate. Rather than use the awkward Toolpak function, determining the residuals involves a relatively straightforward formula, once the regression coefficients are known. But the residuals plot more clearly shows the deviation from the fitted line.

That’s a good point, and something I thought about. The Yankees also have a new stadium, and they’ve had a contending team during the past decade. The Red Sox fall right on the line, but they have a small stadium, and have sold out the last 500+ games in a row.

You would have to look at each season independently, so you could see the effect of a new stadium wearing out, the effect of the previous season’s record, the acquisition of a star player, and other transient effects.

When comparing two variables to look for a possible correlation, then the scatter plot is the most appropriate form of visualization to choose as you point out. A case of “Form follows Function”. (Information Visualization Manifesto).

An excellent chart busters example. The scatter plot is much easier to interpret (and critique) than the bar charts. I wonder whether the correlation coefficient of 39% for the National league is even statistically significant. If not, then even drawing the best fit line is a bit of misinformation.

The Regression tool in the Excel Data Analysis Toolpack has checkboxes to calculate and plot residuals, so it’s easy to look at them.