compare a neighbourhood value with the city mean - test of significance in presence of spatial autocorrelation

Hi everyone,


I have a dataset with all neighbourhoods in a city and a value for each
neighbourhood (a rate). I would like to be able to create a vizualization
that enables the user to select individual neighbourhoods and find out
whether the value of that neighbourhood is significantly higher or lower
than the city average across all neighbourhoods (I have a value that was
calculated for the city). Given that the neighbourhoods are contiguous
areas composing the city, there is spatial dependency, so I cannot use z
scores or apply other classical statistics test since the observations are
not independent and spatial autocorrelation might exist.

I am not sure what test to apply to achieve this. Any help and links to
potential readings would be very much appreciated.

Thank you very much!

Dani

<http://aka.ms/weboutlook>

        [[alternative HTML version deleted]]

Hi Thierry,

Thank you so much! I looked at the BYM model, which I have not used
before. I need to be able to create a table with all the neighbourhoods in
the dataset, indicating whether they are significantly over or under the
city value. I am not sure how to test these differences.

Best regards,
Dani


------------------------------
*From:* Thierry Onkelinx <thierry.onkelinx at inbo.be>
*Sent:* Monday, June 3, 2019 12:23 AM
*To:* dani jones
*Cc:* r-sig-mixed-models at r-project.org
*Subject:* Re: [R-sig-ME] compare a neighbourhood value with the city
mean - test of significance in presence of spatial autocorrelation

Dear Dani,

What I would do is create a graph indicating which neighbourhoods are
directly connected to each other (share a border). And use this information
to fit a Besag-York-Mollier model which takes care of the auto-correlation
among the neighbourhoods. For an example see
library(INLA)
demo("Bym")

You can find the INLA package at http://www.r-inla.org


Best regards,

ir. Thierry Onkelinx
Statisticus / Statistician

Vlaamse Overheid / Government of Flanders
INSTITUUT VOOR NATUUR- EN BOSONDERZOEK / RESEARCH INSTITUTE FOR NATURE AND
FOREST
Team Biometrie & Kwaliteitszorg / Team Biometrics & Quality Assurance
thierry.onkelinx at inbo.be
Havenlaan 88 bus 73, 1000 Brussel
www.inbo.be


///////////////////////////////////////////////////////////////////////////////////////////
To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to say
what the experiment died of. ~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data. ~ Roger Brinner
The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of data.
~ John Tukey

///////////////////////////////////////////////////////////////////////////////////////////


<https://www.inbo.be>


Op ma 3 jun. 2019 om 01:50 schreef dani jones <orchidn at live.com>:

Hi everyone,


I have a dataset with all neighbourhoods in a city and a value for each
neighbourhood (a rate). I would like to be able to create a vizualization
that enables the user to select individual neighbourhoods and find out
whether the value of that neighbourhood is significantly higher or lower
than the city average across all neighbourhoods (I have a value that was
calculated for the city). Given that the neighbourhoods are contiguous
areas composing the city, there is spatial dependency, so I cannot use z
scores or apply other classical statistics test since the observations are
not independent and spatial autocorrelation might exist.

I am not sure what test to apply to achieve this. Any help and links to
potential readings would be very much appreciated.

Thank you very much!

Dani

<http://aka.ms/weboutlook>

        [[alternative HTML version deleted]]