[Openspace] Excellent questions you are asking, Moran's I and Z value question

[Dwight Hines]

Hello, John:

I'm glad that you posted this to the list.  The National Academy of  
Science and USA Today have found tremendous data variability in  
Emergency Rescue and arrivals to the emergency rooms.   In St.  
Augustine, where I'm living, at two different local NAACP community  
meetings, complaints were made about how slow the police and  
ambulances were in responding to calls for help.   I thought it would  
be an easy analysis to get a years worth of call comes time and  
arrival to site of call time and see what differences, if any,  
existed by neighborhoods (St. Augustine, like many southern cities,  
has predominately Afro-American neighborhoods and all white  
neighborhoods.   In case of doubt, the old phone books type a (c) by  
those houses that were Afro-American.

I did not find a relationship between times for the City fire  
department by address or neighborhood, but the city does not provide  
emergency transport.  The county controls the county fire dept and  
emergency medical services.   So far, for the last two years, I've  
been unable to obtain the data, even though the data are not exempt  
from the Florida Public Records Act.   County Sheriff's data are also  
not available for public inspection and copying.

I think with Moran's I  you do the same as you would do with  
Hotelling's T, if there is at a significance you set at a pre- 
determined level, then you go into the data and test for  
interactions, then main effects.   Given the overwhelming data coming  
in from Kaiser, the Robert Woods Foundation, and others on  
disparities in health care by ethnic group, I'd be surprised if you  
did not find significant and substantive differences between northern  
and southern states,  and rural and metropolitan states.  The  
southern states would be confounded by the fact that the Fire rescue  
and emergency services hire primarily white men, while northern  
departments are more representative of the population they serve, but  
not all are so be careful.

Let us know what you are finding and what procedures you are using.    
Also, you might want to check out Freebase <www.freebase.com> as an  
additional method to analyze and display your data for insights or  
patterns that might be missed if you limit your methods to  
traditional statistical approaches.

Note of complete disclosure -- I've never used GeoDa because I'm a  
Mac person.  So, as soon as GeoDa is available for the Mac, I'll use  
it.  Big time.

Dwight Hines,
Independent Media,
Citizen Media Law Project
St. Augustine, Florida










>
> From: "Pearson, John Frederick" <JFPEARSON at PARTNERS.ORG>
> Date: April 13, 2008 10:34:11 PM EDT
> To: <openspace at sal.uiuc.edu>
> Subject: [Openspace] Moran's I and Z value question

> Hello,
>
> I am new to Geoda but enjoying it!  However, I am having a tough time
> understanding how to interpret Moran's I and the z value.  It would  
> be great to
> find an explanation of how important the z value is to a Moran's I  
> score.
>
> I am using Geoda to find the Moran's I for US State level Emergency  
> Department
> visit rate clusters. These rates are the value of ED visits /  
> population.  I am
> also mapping rates of Medically Underserved populations by state  
> and then taking
> these two to do multivariate moran. I am just a little confused on  
> how to go
> about the multivariate and interpret the results properly, with z- 
> value etc.
>
> I can produce the desired results with Geoda (though any tips here  
> would help
> too) but am really looking for some help on understanding them.
>
> Thanks!
> John
>
> John Pearson
>