8.0329 Rs: HOTS; Cluster analysis (2/45)
Elaine Brennan (EDITORS@BROWNVM.BITNET)
Mon, 28 Nov 1994 00:52:51 EST
Humanist Discussion Group, Vol. 8, No. 0329. Monday, 28 Nov 1994.
(1) Date: Fri, 18 Nov 1994 10:45:49 -0500 (EST) (15 lines)
From: "DONALD A. COLEMAN (EXT. 2850)" <DACOLEMAN@FAIR1.BITNET>
Subject: Re: 8.0312 Qs (Part 1): HOTS; TA Training (2/61)
(2) Date: Fri, 18 Nov 1994 15:40:26 GMT (30 lines)
From: lana@rs950.cisi.unito.it (maurizio lana)
Subject: Cluster analysis
(1) --------------------------------------------------------------------
Date: Fri, 18 Nov 1994 10:45:49 -0500 (EST)
From: "DONALD A. COLEMAN (EXT. 2850)" <DACOLEMAN@FAIR1.BITNET>
Subject: Re: 8.0312 Qs (Part 1): HOTS; TA Training (2/61)
on "HOTS:"
I don't know anything about the use of the word "hots," but it does strike me
that, just perhaps, it might be useful to note that "higher order thinking"
might mean thought of a sort that can be brought to bear profitably on matters
which, before they come to light, could not have been predicted.
Anyone who doesn't think that such thought exists doesn't, in my opinion, under-
stand the liberal education at all.
Don Coleman, Asst. Prof. of phil.
(2) --------------------------------------------------------------50----
Date: Fri, 18 Nov 1994 15:40:26 GMT
From: lana@rs950.cisi.unito.it (maurizio lana)
Subject: Cluster analysis
Barbara,
the problem you described is quite common: this is the reason why statistics
and their analyses should be read and accepted carefully ...
=ABWorking with "MacDendro", I experimented with different methods to=
compute
distances and hierarchies, and the results were quite diverging. In fact, I
could find a tree for pretty much any point I could want to make.=BB The=
fact
is that if you submit 30 items to the program and ask: make a hierarchy of
them, the program can't do anything but insert *all* those items into the
hierarchy! it can't at all "decide" that -say- 4 items belongs to a
completely different tree; it can only show that a certain branch of the
tree is very different and very apart of the tree with the other branches.
So the researcher must read the trees and give an interpretation of them,
discarding those which are wrong, and describing those (or that one) that
are (is) interesting.
Or, conversely: the many hierarchies you get, show that the data you are
studying are much more complex than one could have supposed.
I understand that perhaps my poor english isn't the best tool to describe
these subtle matters, but hope I managed to al least give you some hints.
Best wishes
Maurizio
Maurizio Lana - CISI Universita'di Torino
Via S. Ottavio 20, Torino - Italy
fax 39 11 8991648