lims <- quantile(sdcontrib, probs = c(.005, .995)) # obtain 1%-extreme data
is_normal <- which((sdcontrib > lims[1]) & (sdcontrib < lims[2]))
-matcontri_ext <- matcontrib0[-is_normal, ]""
+matcontri_ext <- matcontrib0[-is_normal, ]#""
matcontrib <- matcontrib0[is_normal, ] # wipe out aberrant data
matcontrib <- t(apply(matcontrib, 1, function(x) x / sum(x)))
## c. Clustering ##########
K <- 200
system.time(
+
+#TODO: cette partie en C
+
clfit <- clara(x = tdata[, selvar],
k = K,
sampsize = 4000,