[R-sig-Geo] Moran I with Edge effect correction

[Roger Bivand]

On Mon, 27 Aug 2012, laurence delattre wrote:

> Hi,
>   This is very interesting. I would like to know a little bit more about
> how to use glist=argument to reduce weights on observations selected as
> members of an edge set and how to identify such members.

There isn't a simple answer, so each case would need to be guided by the 
analyst. If a concave hull is available, one could down-weight all links 
to observations in the set of members of the concave hull. If the 
observations have polygon support, one could mark all polygons with less 
that 100% internal shared boundaries (having an external boundary) for 
down-weighting. The down-weighting might be done by iterating over the 
links in the nb object to apply the downweighting factor to each link. 
There is a discussion of concave hulls on:

http://r.789695.n4.nabble.com/Concave-hull-td863710.html

example(nc.sids)
nn <- card(ncCR85_nb)
summary(nn)
nc.sids$nn <- factor(nn)
spplot(nc.sids, "nn", col.regions=rev(grey.colors(9)))


But your choice of weights style will have a major impact, so for binary 
weights, the low neighbour count observations are already down-weighted. 
The problem of up-weighted low neighbour count observations is most 
pressing for row-standardised weights, and there the glist= argument will 
not help. These topics are not well covered in the literature, I'm afraid.

Hope this helps,

Roger



>  Thanks in advance.
> Laurence
>
> 2012/8/25 Roger Bivand <Roger.Bivand at nhh.no>
>
>> On Thu, 23 Aug 2012, Kitty Lee wrote:
>>
>>  Hi. I have some study areas that are very irregularly shaped. I wonder if
>>> there is a way to adjust for edge effect in computing Moran's I or Geary C?
>>>
>>
>> The typical approach would be through the weights matrix, depending on how
>> W is constructed. If you row-standardise, edge observations will be
>> stressed, but with binary weights they are less salient. You could also
>> intervene directly to reduce weights on observations selected as members of
>> an edge set, through the glist= argument in nb2listw().
>>
>> Hope this helps,
>>
>> Roger
>>
>>
>>
>>> Any leads would be really appreciated.
>>>
>>> K.
>>>
>>> ______________________________**_________________
>>> R-sig-Geo mailing list
>>> R-sig-Geo at r-project.org
>>> https://stat.ethz.ch/mailman/**listinfo/r-sig-geo<https://stat.ethz.ch/mailman/listinfo/r-sig-geo>
>>>
>>>
>> --
>> Roger Bivand
>> Department of Economics, NHH Norwegian School of Economics,
>> Helleveien 30, N-5045 Bergen, Norway.
>> voice: +47 55 95 93 55; fax +47 55 95 95 43
>> e-mail: Roger.Bivand at nhh.no
>>
>>
>> ______________________________**_________________
>> R-sig-Geo mailing list
>> R-sig-Geo at r-project.org
>> https://stat.ethz.ch/mailman/**listinfo/r-sig-geo<https://stat.ethz.ch/mailman/listinfo/r-sig-geo>
>>
>
>
>
>

-- 
Roger Bivand
Department of Economics, NHH Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no