[Openspace] Randomization for a subset of data

[Julia Koschinsky]

Hi Emi,

The pseudo p-values are currently only computed for the 
whole dataset, not subsets. If you wanted to know if subsets 
are independently significant, you could use a workaround of 
creating separate subsets after you identify relevant 
subsets through brushing. This workaround is clumsy and not 
on-the fly and you have to work with separate subsets of the 
data instead of keeping your total N.

However, if you have somewhat distinct subsets of N you want 
to test, you could select the observations of interest, then 
right-click on the table, go to "Save Sel Obs," which allows 
you to add a dummy for the selected obs, and save as new 
shape file (if you have multiple subsets, do this multiple 
times). Then create new subsetted shape files in a GIS (or 
edit the shape file text file manually after exporting it), 
bring the subset files sequentially back into GeoDa and run 
the test statistic again.

Let me know if any of the details are unclear.
Julia

---- Original message ----
>Date: Wed, 21 Jun 2006 16:56:14 -0700
>From: "Emi Uchida" <euchida at gmail.com>  
>Subject: [Openspace] Randomization for a subset of data  
>To: openspace at sal.uiuc.edu
>
>I am using "brushing" to calculate global Moran's I for 
selected
>observations.  I can see Moran's I's statistics, but is 
there a way to
>perform randomization just for the selected observations to 
get the
>p-value?
>
>If not are there other ways to select a subset of 
observations and calculate
>p-value for Moran's I or perform other tools in GeoDa?
>
>Thanks,
> Emi
>_______________________________________________
>Openspace mailing list
>Openspace at sal.uiuc.edu
>http://sal.uiuc.edu/mailman/listinfo/openspace