Over here: http://www.businessi...america-2013-6#
Nice use of colour to present the data.
Jump to content
Except that it masks the sample size. From these maps you might assume that hundreds of thousands of people were surveyed, instead of 10,000. On the original maps you can see the areas of the country that weren't heard from at all.
Maps should not imply false precision.
Sample size is very improtant and perhaps they should have a clearer indication of the sample in the final maps, but I'd say they pretty much nailed the US population distribution:
The purpose of maps like this are to reveal patterns and I think they got decent results. Some of the patterns are more subtle than others and may have improved with a larger sample size, but I don't feel lied to. Maybe we have an advantage other map readers don't in understanding how maps like this are built but I don't know if I'd suggest it was bad science or bad mapping. It's actually a pretty typical methodology.
I wouldn't look at this as a predictive surface like you might a precipitation or tempertaure map, although both use sampleing and interpolation to extend limited samples into full coverage surfaces.
For comparison NOAA says they currently rely on only 1500 weather stations (down from 6000) to model global climate. NWS has around 2000 stations (plus a 5000 station "cooperative" network) in the US. Pretty similar, and ultiamtely smaller, sample sizes.
GIS Reference and Instruction Specialist, Stanford Geospatial Center.
Fascinating maps! I am glad Dennis found the raw data as I was wondering about their sample size as well. It would be interesting to see the ratio of the sample size to actual population by state to see if there was any implied skew.
I think it was rather well done and mirrored my own dialect (eg. I say crick and was raised in Montana).
Screen shot 2013-06-06 at 3.10.40 PM.png 50.32K 1 downloads
Oregon Metro - Portland, OR
0 members, 0 guests, 0 anonymous users