5 ms·
The solution in the article was a simple scatterplot of the data points. There's no need to use any boundaries when just examining the data. The basically unbe
by maxerickson 10y ago
The solution in the article was a simple scatterplot of the data points.
There's no need to use any boundaries when just examining the data. The basically unbelievable thing that happened is that they only looked at one summary of the data.
- tyingq 10y agoI agree with your last point. A scatter plot, though, could just reflect population density. There is still value in bounds so you can see relative figures. I don't see a silver bullet really. The difference here seemed to be having someone with the right skills.
- maxerickson 10y agoSure, it wouldn't be the end of the analysis or necessarily the thing you'd use to communicate the data. But here's a paragraph from the article about the distribution of the data: As I ran the addresses through a precise parcel-level geocoding process and visually inspected individual blood lead levels, I was immediately struck by the disparity in the spatial pattern. It was obvious Flint children had become far more likely than out-county children to experience elevated blood lead when compared to two years prior. "It was obvious"...
- tantalor 10y agohttps://xkcd.com/1138/ https://xkcd.com/1138/
- maxerickson 10y agoAnd still the geographer that wrote the article has the temerity to suggest that the pattern was obvious in a simple plot of the data.
- tantalor 10y agoYes, I guess you need 2 scatter plots; 2 years ago and now. (assuming population density is same in both)
- maxerickson 10y agoI think it is simpler than that. Baseline lead exposure is very low so absent a specific problem high density areas will still have few high test results, and the problems with accidentally making a density map aren't so prevalent when a small area is under consideration.
- tyingq 10y agoIn this case it was. But, water isn't the only way to get lead exposure. There were the novelty drinking glasses distributed by McDonald's for example. Or kids eating lead based paint flakes from the porches of older multi-tenant buildings. Basically just trying to say that there's value in having experts around versus thinking one type of methodology just always works. As I understand it, in this case, the water supply wasn't contaminated with lead. It was either more acid or more alkaline than it used to be, which caused the individual houses with lead pipes to leech more lead. That's a pretty specific causality where a scatter plot was the most useful approach.
- maxerickson 10y agoYeah, it was the interaction between the water and pipes. Here's the baseline data though: https://docs.google.com/spreadsheets/d/1driDoWu_lz1IHf1NHVFkc7MT3nr1G0luaoaCgTJSn5o/edit#gid=139983916 https://docs.google.com/spreadsheets/d/1driDoWu_lz1IHf1NHVFk... In most counties in Michigan, there are very few addresses with a child that has been tested and has high blood lead levels. So even a small cluster prompts further investigation.