how to distribute ZIP level data to tract or block or arbitrary polygon?

Question

I can't crack a problem I've been working on for quite a while.

I have ZIP-level data for a certain thing, let's say hospital discharges or ED visits.

I want to "dub down" this ZIP-level data to tracts or blocks. This data scales better by population, not area.

Summarize within only calculates by area, right?  How can I do this in AGOL?

As we all know, some ZIPs are very "lumpy" in where actual people live, especially out in the west where many ZIPs are mostly empty space. So tracts and blocks show were actually the people are.

I want my output to be the tracts/blocks/or polygon enriched with the correct number of discharges scaled by population.

Imagine we had a ZIP that was a perfect 10mi by 10mi square.  100 sq mi. 1000 pop and 1000 discharges.

Imagine there's 2 blocks in there that are b2 =10mi by 7mi (70 sq mi) and b1 = 10mi by 3mi (30 sq mi). b2 has almost nobody, only 100 pop, while b1 has pop 900.

I want the 1000 discharges for 1000 pop at ZIP level distributed or "enriched" to the blocks b1 and b2 by pop, resulting in b1 = 900 discharges and b2 = 1000 discharges.  If I use summarize within, it goes by area, and gives all the discharges to b2 because it is bigger.

How do I do this in AGOL?

JayLarsen · Answer

Maybe I can explain it better?

I have ZIP codes with known numbers of population and discharges.

I want to "dub down" the discharge data to census tracts or blocks (or even arbitrary polygons such as a tessellation).

Step 1 would be to ENRICH the tracts/blocks/whatever with population data.

Step 2 would be to find a function that can relate the boundaries of the ZIP pop and discharges to the boundaries of the tract/block whatever by POPULATION.

So, if a big ZIP had 100 discharges and 100,000 people (.001 discharges per person) and it had a tract that was half the size of the ZIP, but only had pop 100, that it would distribute only .1 discharges to the tract.

Does that help? The FIND EXISTING function only uses area when "distributing" a property.

Essentially, I want to ENRICH my tracts or blocks using the existing discharge data and the known populations, not areas.

JayLarsen · Answer

Yeah, you're not quite getting my problem.

Trust me, there will be discharges all over a metro area no matter how far a given hospital is.

The key is to distribute the known discharges in a given ZIP code down to the census tracts/blocks or even arbitrary shape by POPULATION in that block, not by AREA of that block.

Again, we have areas with almost no population that are large, so using SIZE to distribute would be silly.

Any other ideas?

JanetBrewster1 · Answer

Hi Jay,

Not even getting to the "how to", my initial thought is that this can't be done accurately without knowing where the hospitals are and the data per hospital -- assuming that's what you're after, as opposed to where the people LIVE who were discharged from any given hospital.

In your example of blocks b1 and b2, what if b2 doesn't contain any hospitals? There would be no discharges in that block, so it's not a simple algebraic calculation. And unfortunately, zips and census boundaries don't align.

Or maybe I'm completely misinterpreting what you're after....

Janet