Fri, June 11, 2004

The Daring and magnificent John Gruber has been talking a bit about security on the Mac platform lately, because of the recent publicity (some deserved, some not) that it's been getting. All of it well worth reading, even if you don't have a Mac (or a computer at all, which I guess means you're at the library right now). Particularly Security Cannot Be Spun.

Today, he posted this little chart of market share as measured by Google. It's quite interesting, but I think the point should be made that this is a slightly skewed sample.

This isn't really computer market share, it's computer market share as represented by computers that do a Google search. (I'm guessing this includes Google-based searches on other sites, but I'm not sure.) The point is, they're on the net, and they're visiting Google. If they're on the net, there's a better than average chance they're at the office, and most offices are outfitted with PCs, even if the users might have Macs at home. (However, I'm sure a higher percentage of Mac owners are connected than Windows users, simply because it's a smaller pool and the income and computer use skew slightly higher in the Mac group.) So, I can't say which way the data is skewing, but it's always important to consider the source and the conditions for testing.

Same with this:

Pop vs. Soda by County

Are these mall interviews? Online polls? What qualifies a respondent for a specific county? Do people supply their zip codes or is it just the county in which the poll took place?

In our current USA Today world, I think analytical thinking is woefully lacking when approaching statistical data. Not that I'm accusing Gruber of sloppy analysis. He's the best there is. It's a perfectly good measure of market share, it's just not the same as "percentage of every computer owned by anyone." It's close, and much easier to attain. The thought just popped in my head when looking at the chart, and the Coke map. Maybe I'm channelling Edward Tufte here.