cut classification by within-group Chi-square: experimental code

Rolecek et al (2009) modified twinspan so that instead dichotomic
splits by level it will only split the most heterogeneous group
at each step. This function tries to achieve the same by returning
a classification vector so that most heterogeneous classes are split.

R/cut.twinspan.R

@@ -54,3 +54,31 @@
     }
     cl
 }
+
+## cut by group homogeneity as defined by within-group
+## Chi-square. This could give similar clustering as Rolecek's
+## modified twinspan which only splits the most heterogeneous group at
+## each step instead of dichotomizing. However, we use the internal
+## twinspan criterion of Chi-square instead of dissimilarities. There
+## is no guarantee that the tree will be ordered, but let's hope so
+## and as.dendrogram(x, height="chi") will show this.
+
+#' @export
+`chicut` <-
+    function(x, ngroups)
+{
+    if (missing(ngroups))
+        ngroups <- 1 # return one group if nothing is asked
+    chi <- twintotalchi(x)
+    chi[chi <= 0] <- NA # not evaluated
+    ix <- order(chi, decreasing = TRUE) # order by heterogeneity
+    clim <- 2^(0:x$levelmax) - 1L
+    class <- rep(1, x$nquadrat)
+    for(i in seq_len(ngroups - 1)) {
+        lev <- max(which(ix[i] > clim))
+        id <- 2L * ix[i]
+        class[cut(tw, level=lev) == id] <- id
+        class[cut(tw, level=lev) == id+1L] <- id + 1L
+    }
+    class
+}