[R] Spatstat - Several density plots using the same scale

[Adrian Baddeley]

Arthur Weiss wrote:
>
>> Hi everyone,
>>
>> I am using the package "spatstat" for ploting kernel maps of my data.
>> It is a marked point pattern, the result of mosquito surveillance in a
>> area in a week.
>> For each trap, the number of individuals captured is the mark of the 
>> point.
>>
>>> plot(density(X, weights=X$marks))
>>
>> makes a nice kernel, but the problem is that I've got several weeks and
>> for each week the density is re-scaled, which makes comparisons useless.
>> I've been trying to find some way to pass the scale limits to the 
>> function
>> but just couldn't find.

See help(plot.im) for information about how to plot a pixel image in 
spatstat. The argument 'zlim' controls the range of numerical values 
that are mapped to colours in the display.

If you have, say, 4 pixel images Z1, Z2, Z3, Z4 and want to plot them 
all with the same colour map, try the following.

    Zlist <- list(week1=Z1, week2=Z2, week3=Z3, week4=Z4)
    Zrange <- range(unlist(
                         lapply(Zlist, function(x){summary(x)$range})))
    plot(as.listof(Zlist), zlim=Zrange, ncols=2)

However, it's not clear that the command density() is really what you 
want to use in this context. This command estimates the 
spatially-varying average intensity (`density') of points. Are the 
insect traps at fixed locations that were chosen by the experimenter? If 
so, then it is somewhat meaningless to estimate their average 
density.... What you need is a method for spatially interpolating the 
insect counts (number of trapped insects) observed at these locations.  
In spatstat you can use the command 'smooth.ppp' to perform 
kernel-weighted spatial interpolation. If the insect counts are small, 
then it would be more appropriate to do spatial Poisson regression 
(using other packages).

Adrian Baddeley