Missing data in linkage trees

Note that LINKTREE is able to tolerate some missing data in the abiotic matrix – the piecemeal form of LINKTREE’s conclusions lends itself to analysing what¬ever complete matrix is available locally, i.e. within each created subdivision. But distortions in interpretation from unavailability of explanatory variables in some sets of samples and not others are almost inevitable. A final point to make is that it is always interesting to compare a constrained Cluster>LINKTREE with the un-constrained, but otherwise very similar, Cluster>UNCTREE tree structure. Here, exactly the same divisions are found (and of course confirmed in the same way by the SIMPROF tests). Where there are major differences, this suggests that natural clusters in the samples are not being well identified by the current abiotic suite, perhaps because a key variable is missing (though there are many other possible reasons! – see the discussion on reversals in B% plots in Chapter 11 of CiMC).