3 ms·

Always disappointing to see counties used for comparison. Also disappointing to see totals instead of per-capita values.

arjie 7y ago
What would you prefer?
- kodablah 7y ago
  They make statistical areas[0], granted they can be seen as too large to be meaningful. Rather, in this case, to determine flow, you have to use adjacency to determine "areas" of large concentration that may consist of a single county or multiple. A hard line on a county border can give inaccurate results. 0 - https://en.wikipedia.org/wiki/Statistical_area_(United_States) https://en.wikipedia.org/wiki/Statistical_area_(United_State...
bl4ckneon 7y ago
Per-capita doesn't make sense in this case. This is about the total flow of food. It again wouldn't make sense to say that a small town in the middle of no where with say 100 people who have the highest per-capita food export when in total they aren't even a drop in the bucket in importance to somewhere in CA or the Midwest that is an agriculture power house and accounts for most of our total food production.
- kodablah 7y ago
  But total flow is skewed when one county is an order of magnitude bigger than another in size. If you are going to compare counties which are not a fixed physical size, you at least need to account for differing physical size's effects on population or you can't reasonably compare. Were the counties a quarter of their current physical size, capita split and all, would they still appear as significant?
dredmorbius 7y ago
Much of this depends on data availability. I'd need to dig into the paper, but depending on reliance, USDA or SIC classifications by county (ZIP Code is also often available). Data availability tends to drive much analysis and research.
- kodablah 7y ago
  Sure, and if you only have county-level metrics, then you can only perform analysis at the county level. But then you must be aware of the skew that represents in data, and make that clear. Presenting county-by-county analysis as definitive representations of flow, not accounting for physical size, does not give an accurate picture of real flow. Only to population centers and/or large counties, but large flow may be happening to many adjacent, small physical counties and this type of analysis of "core counties" would never know.
  - dredmorbius 7y ago
    A serendipitous advantage of the cookie-cutter rectangular definition of many ag states' county boundaries is that regions are fairly strongly normalised for land area. And ag productivity, for some strange reason, correlates strongly to area rather than population, all else equal. California would be an exception, though the distortions aren't particularly exceptional. A rough eyeballing of the county-based graph (lower in my Mastodon reshare) suggests if anything that larger counties are less rather than more represented in high-significance areas. Most large counties tend to be both exceptionally sparsely populated, and agriculturally marginal. San Bernardino County, in California, is a notable exception.

arjie 7y ago

What would you prefer?

kodablah 7y ago
They make statistical areas[0], granted they can be seen as too large to be meaningful. Rather, in this case, to determine flow, you have to use adjacency to determine "areas" of large concentration that may consist of a single county or multiple. A hard line on a county border can give inaccurate results. 0 - https://en.wikipedia.org/wiki/Statistical_area_(United_States) https://en.wikipedia.org/wiki/Statistical_area_(United_State...

bl4ckneon 7y ago

Per-capita doesn't make sense in this case. This is about the total flow of food. It again wouldn't make sense to say that a small town in the middle of no where with say 100 people who have the highest per-capita food export when in total they aren't even a drop in the bucket in importance to somewhere in CA or the Midwest that is an agriculture power house and accounts for most of our total food production.

kodablah 7y ago
But total flow is skewed when one county is an order of magnitude bigger than another in size. If you are going to compare counties which are not a fixed physical size, you at least need to account for differing physical size's effects on population or you can't reasonably compare. Were the counties a quarter of their current physical size, capita split and all, would they still appear as significant?

dredmorbius 7y ago

Much of this depends on data availability. I'd need to dig into the paper, but depending on reliance, USDA or SIC classifications by county (ZIP Code is also often available). Data availability tends to drive much analysis and research.

kodablah 7y ago
Sure, and if you only have county-level metrics, then you can only perform analysis at the county level. But then you must be aware of the skew that represents in data, and make that clear. Presenting county-by-county analysis as definitive representations of flow, not accounting for physical size, does not give an accurate picture of real flow. Only to population centers and/or large counties, but large flow may be happening to many adjacent, small physical counties and this type of analysis of "core counties" would never know.
- dredmorbius 7y ago
  A serendipitous advantage of the cookie-cutter rectangular definition of many ag states' county boundaries is that regions are fairly strongly normalised for land area. And ag productivity, for some strange reason, correlates strongly to area rather than population, all else equal. California would be an exception, though the distortions aren't particularly exceptional. A rough eyeballing of the county-based graph (lower in my Mastodon reshare) suggests if anything that larger counties are less rather than more represented in high-significance areas. Most large counties tend to be both exceptionally sparsely populated, and agriculturally marginal. San Bernardino County, in California, is a notable exception.