[S] spatial.cor, binary

[Jane Elith <j.elith@botany.unimelb.edu.au>]

 Dear Splus,
A colleague has a question about the computation of a 2-outcome spatial autocorrelation index using the spatial.cor function in S+ Spatial stats. At present it appears the binary option has been withdrawn, leaving only the Moran's i and Geary's c options available. Neither of these options appear to be suited to binary outcome data. If you have experience with this function or know how implement it with binary outcome data could you email him directly?:

Brendan Wintle
b.wintle1@pgrad.unimelb.edu.au


(we work with S+ v4.5 on Win NT or Win 95, and Spatial Stats v1.1)

Here is his request:
I've been trying to compute the spatial correlation index for binary
(presence/absence) data using the spatial.cor function in spatial  stats.
The manual I have (v1, 1996) tells me that I can use the option;
statistic="binary"(pg D-71).  When I use this option (on v1.1 1999), it
returns Error; 'statistic "binary" not implemented in this version'.  

Can you please tell me why this option was withdrawn? If I use the
'statistic="moran" option on a binary outcome set in spatial stats, is the
computed result equivalent to the binary measure of spatial  association?
If not, do I have any other options for computing Moran's (1948)
two-outcome measure of spatial association. 


Here is part of an answer from Mathsoft:
>>>>
> I found an existing bug report saying that the "spatial.cor()" function
> does not do statistic="binary".  I also found the following discussion in
> our database about "binary" option for "spatial.cor()" function:
> -------------------------------------------------------------------------
> QUESTION:
> > Is there anything in S-PLUS for modelling binary spatial data, or is
> > there planned to be?
>
> ANSWER:
> > I discovered that the binary statistic for spatial.cor was not  correct
> > just before we released SpatialStats.
> >
> > We intend to put the binary statistic back into spatial.cor for the
> > next release of S+SpatialStats (no set date for that yet).
> -------------------------------------------------------------------------
>
> I checked the newest version of S+SpatialStats here and see that we
> haven't put the "binary" statistic back into "spatial.cor()" function yet.
> Thus, according to the above, we currently do not have any options that
> deal with binary spatial data.
<<<<


We appreaciate any help,
Jane Elith

----------------------------------------------------------------------- This message was distributed by s-news@wubios.wustl.edu.  To unsubscribe send e-mail to s-news-request@wubios.wustl.edu with the BODY of the message:  unsubscribe s-news