Hacker News Reader

Comment by mixologic | Hacker News Reader

Wow. That is the worst case I've ever seen of map analysis.

Go to the original map here https://www.mapbox.com/labs/twitter-gnip/brands/#, and turn off the iphone layer that happens to be on top. This is pathetic.

Im gonna call my buddy at mapbox.

View on HN

Yeah, this is pretty egregious. When I played around with the source map and Manhattan, most of the effect they mentioned goes away when you look at the maps in isolation. The iPhone layer was drawn on top of Android, so of course it looks like Manhattan is full of iPhone users; the only time the Android pixels stand out is where there are no iPhone users at all.

Even with the egregious layering problem, the map actually shows something pretty interesting: the way in which Android has colonised lower-density lower-income areas to the almost total exclusion of the iPhone. It makes sense: you can buy brand-new low-end Android devices; you can't do the same with iPhones.

So the map doesn't actually show that Android is doing poorly in high-value areas; it does, however, show that Android is doing well in low-value areas -- in addition to doing well in high-value areas. The data is a win for Android even if the map literally obscures this fact.

You're totally right. Example: Manhattan has a higher concentration of Android than Newark by far. This is not visible without toggling layers. This is probably due to the concentration of people in Manhattan vs. Newark, IMHO.

The argument that iOS is more frequent in high income areas still seems fairly sound. Still, the story misrepresents the map by far.

It is bad analysis by the reporter but it's also a bad map.

Once the tweet-density becomes higher than the maximum point density the area turns red. So any dense concentration of people is shown as red. Find a crowded poor neighborhood and it'll be red. And it's compounded by the way red pops out more than the green does. Also compounded by the annoying zoom level restriction.

The maps aren't great but Manhattan seems like a particularly bad example due to the point density. If you look at Chicago (https://www.mapbox.com/labs/twitter-gnip/brands/#11/41.8726/...) the distribution of phone types with regard to the economic status of the neighborhoods is a little clearer. Obviously it's all basic/easily misled visual comparison though.

What we can infer is that people with iphones don't tweet in poor neighborhoods

Please let us know what your buddy at Mapbox says.

"facepalm"

For the advanced student: this is a good comment for HN. Its signal/noise ratio is very high, given its context. There's nothing wrong with brevity.

It's true that one-word comments where the word is facepalm would nearly always deserve downvoting. Discernment is required!

Facepalm all by itself would deserve downvotes. It's the quotation marks that set this apart.

Exactly the kind of discernment we need.

Also android seems to be the bottom layer no matter what order i turn the layers on. This is idiotic.

Or a really clever marketing tactic:

Hmm, do I want a phone associated with rich people, or poor people?

The article's map data almost seems unrealistic to me. Something is way off...

Reply on news.ycombinator.com