R: Multi-type Point Pattern scan.test

This is not supported, partly because it is problematic from a statistical viewpoint.

First, do you want to compare all pairs of possible types (i, j) or compare each type i against 'all other types' (not-i) ?

From a statistical viewpoint, this suffers from the "multiple testing problem"; since you are performing several hypothesis tests, the chance of getting a spuriously 'significant' outcome is much greater.

From a software viewpoint, the results of these multiple tests should be returned as a list of hypothesis tests (a list of objects of class htest), which would make it incompatible with the output of other uses of scan.test which return a single object of class htest.

You can do these pairwise tests using a loop if you really want to.

There is a multinomial (i.e. generalisation of binomial to more than 2 types) version of the scan test. This would probably be the "right" way to treat these data. The complication there is to specify the alternative hypothesis (e.g. do you want to detect all instances where the pattern of cases is significantly different from the baseline? or only those instances where type 1 is significantly more abundant than types 2 and 3 relative to the baseline? etc etc) So this is not yet implemented either.