Bob,  Caliper will have to give you an answer, but I do this type of operation a lot. Of course this is important to do this as in my state, anyway, block groups have changed a lot since 1990 (in some areas)  That is why you are doing an overlay? You should be able to get perfection in the overlay process.
 
From a statistical view, you might get more reliable results with census tracts. Even bigger geographies like neighbors or communities. Even Zip Codes.
 
If you are looking for trends using any SF3 data, census survey data, you can get some screwy answers and likely you will not know it as the geography gets smaller.
 
The total ascertainment variables, SF1, should be OK at least for a simple trend, but be careful as just because you get a difference, this does not mean there really is one even if the block groups exactly overlay. When you are doing areal overlay there is added variability because of the assumption that the block group has a homogenous distribution of the population variables. So you have two sources of variability, the variability due to survey variance (not easy to calculate) and the variability in the overlay progress.
 
In order to calculate confidence intervals or do a statistical test, you really have to do a Monte Carlo simulation as the distributions are not normal, and the numbers are not robust with respect to assuming that they are. They become Poisson or gamma distributed. Simulation because there is no closed form formula for calculating the confidence intervals.
 
Bottom line, I guess you have to not care too much about perfection but also take the results with a grain of salt.
 
Richard Hoskins
WA State Dept of Health


From: [email protected] [mailto:[EMAIL PROTECTED] On Behalf Of [EMAIL PROTECTED]
Sent: Wednesday, February 01, 2006 10:56 AM
To: [email protected]
Subject: [Maptitude] Overlay woes

I'm trying to overlay the 1990 block groups onto the 2000 block
groups. These are the geography/data cds that I bought from Caliper.

What happens is the overlay starts to process. and then it stops with
the message "Encountered topological problems when processing
polygons 14708142 and 18188130. Error at or near coordianate (-
119571922,364891287) Click OK to continue. Reference info:
aggr,321,0."

My goal is to calculate some changes in demographics. Like the growth
in hispanic population from 1990 to 2000.

When I zoom to that area (it's in the 1990 layer), I can see that
these polygons overlap each other.

Is there some kind of workaround or fix? I don't need perfection,
just trends.

Bob


YAHOO! GROUPS LINKS




Reply via email to