[R-sig-Geo] spplot: labels on maps / variables on different scales

[Roger Bivand]

On Tue, 26 Aug 2008, Michael Friendly wrote:

> Edzer Pebesma wrote:
>> 
>>
>>  Michael Friendly wrote:
>> >  Two short questions about working with maps:
>> > 
>> >  1.  I'm reading a shapefile with character labels for the regions (FSA). 
>> >  I can add the labels using plot(),
>> >  but when I try the same thing using spplot(), the labels are in the 
>> >  wrong positions -- they all seem to be
>> >  shrunk somewhat in toward the center of the map.  What am I doing wrong?
>> > 
>> >  # this doesn't work-- labels in wrong position
>> >  spplot(toronto,"FSA_NAME", colorkey=FALSE)
>> >  text(coordinates(toronto), labels=as.character(toronto$FSA), cex=0.4)
>>  Right: text() works with base graphics, not with lattice on which spplot
>>  is built.
>>
>>  Something like this should work:
>>  spplot(toronto,"FSA_NAME", colorkey=FALSE,
>>     sp.layout = list("sp.text", coordinates(toronto),
>>  as.character(toronto$FSA), cex=0.4))
>> 
> Great!  Now I also know where to look to generalize this.
>> > 
>> >  2. I have a bunch of attribute variables for the geographic regions, all 
>> >  on different scales.  Id like to
>> >  produce a set of comparative maps in the same figure (say with spplot()) 
>> >  with each attribute shaded
>> >  by its quantiles, e.g., 5 classes each.  Do I have to precompute these 
>> >  first, or is there something I can do in the call
>> >  to spplot() to have this done, using the variables in the 
>> >  SpatialPolygonsDataFrame?
>>  What exactly did you mean by "all on different scales"? They have
>>  different polygon structures?
> No - some of the attribute values are percents, some are quantitative & 
> positively skewed, like Income. If I do
>
> spplot(toronto, c("Household.Income","Unemployed","University"))
> a single scale is applied to all three, so the two % variables are shaded 
> uniformly in the lowest range.
> What I'd like is to apply a function to take each of these and recode into 
> quantiles for that variable.

After a little digging around, it looks as though the plot() method for 
trellis objects (pp. 202-206 in the Lattice book) provides a way to 
generate a single graphic from multiple calls to spplot, something like:

p1 <- spplot(toronto, c("Household.Income"))
p2 <- spplot(toronto, c("Unemployed"))
p3 <- spplot(toronto, c("University"))
plot(p1, split=c(1,1,2,2), more=TRUE)
plot(p2, split=c(1,2,2,2), more=TRUE)
plot(p3, split=c(2,1,2,2), more=FALSE)

using at=, col.regions=, main=, etc. in each of the spplot calls as 
appropriate for the selected variables. With the same col.regions= and at= 
based on quantiles (perhaps floor() for the first and ceiling() for the 
last), this should be pretty close visually, but with a key for each 
variable.

>
> It's partly that my data variables are now in the map object and, from the 
> help, I only know how to refer to
> zcol= names of these, rather than some transformations on the underlying 
> data.

The alternative might be to assign new derived variables to the 
Spatial*DatatFrame object, which for all intents and purposes "is" a data 
frame, and spplot() them.

Roger

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, 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