>a) why can COA accept indicator matrix data? (This confuses me)
A table with 0 and 1 can be a so called "complete disjunctive table",
and in this case it is equivalent to a qualitative variables table.
The columns are indicators of the qualitative variables and it can
be called an indicator matrix.
But a table with 0 and 1 can also be a presence/absence table.
In the first case, you can use the COA module to perform the analysis of
this complete disjonctive table. This is exactly the same thing as the
multiple correspondence analysis of the qualitative variables table.
It is generally preferable to use the MCA module directly on the qualitative
variables table rather than first computing the complete disjunctive table
(for example with CategVar : Categ->Disj) and then using the COA module,
but both ways give the same results.
In the second case (presence/absence), you can use the COA module to compute
the simple correspondence analysis of the presence/absence table. You cannot
use the MCA module, which can only be used on qualitative variables tables
(not on complete disjunctive tables, and not on presence/absence tables).
See the message of Daniel Chessel (MCA_COA_etc) for more details, references,
and Burt tables (on which you can also use COA).
Also note that in ADE-4, a 0/1 table cannot be a qualitative variable table
(with 0 = first category and 1 = second category). The categories of a
qualitative variable must always be numbered from 1 to m (m = number of
Additional question : are there relationships between the COA of a presence/
absence table and the MCA of the corresponding qualitative table (1 = absence,
2 = presence) ?
-- Jean Thioulouse - Laboratoire de Biometrie - Universite Lyon 1 69622 Villeurbanne Cedex - France Fax: 04 78 89 27 19 Tel: 04 72 43 29 01 http://biomserv.univ-lyon1.fr/JTHome.html
This archive was generated by hypermail 2b30 : Sat Feb 10 2001 - 10:21:43 MET