+------------------------------------------------------+
 | Correspondence Analysis with SAS Enterprise Guide v2 |
 +------------------------------------------------------+
 |                                                      |
 | Custom Task for SAS Enterprise Guide which offers a  |
 | minimal GUI wrapper for SAS CORRESP procedure, thus  |
 | allowing to perform Simple Correspondence Analysis   |
 | (SCA) or Multiple Correspondence Analysis (MCA).     |
 |                                                      |
 +------------------------------------------------------+

  Contents of this file
     1   Installation procedure
     2   How to use / Tips
     3   Known bugs / Limitations
     4   Copyrights
     5   Contact the author

 +------------------------------------------------------+ 


 1 Installation procedure


 Copy the file SAScorsp.DLL in the folder where you
 installed SAS Enterprise Guide. 
 
 Register the DLL: open a command windows (cmd)
 Go to the directory where SAS EG is installed (cd)
 And execute: regsvr32.exe SAScorsp.dll
 
 Then Either
	 Launch Enterprise Guide and open the menu Tools|Customize

	 Go to the Add-Ins tab and click the Add... button

	 Enter the ID associated with this DLL:  PROCcorresp.Corresp

	 Restart Enterprise Guide.

 Or execute the file ProcCorresp.reg 

 The method Correspondence Analysis should have been added 
 in the Multivariate category.
 
 
 +------------------------------------------------------+  
 
 2 How to use
 
 
 Correspondence Analysis is a method to find a low-dimensional
 representation of both rows and columns of a crosstabulation
 or contigency table.
 When using Correspondence Analysis within SAS system, you can
 use two kinds of data: raw categorical responses on two or more
 classification variables and a existing two-way contingency 
 table.
 
 First, you have to precise the kind of data that you will use
 (which kind of structure: raw data or contingeny table).
 Then, you assigns the variables.
 
 In order to have a variable play a supplementary role, you have
 to put it also as active variable (supplementary variables are
 projected on the subspace generated by the other active variable,
 thus playing no role to the construction of this space). 
 
 To work with supplementary individuals, you have to specify a 
 weight variable. Individuals with negative weights will be removed 
 from the analysis and treated as supplementary. You can build the
 adequate weight variable prior the analysis in a datastep such as:
 
	DATA mylib.mymember;
 		SET mylib.mymember;
 		IF _N_=1 or _N_=17 or IDvar="label"
 		THEN w=-1;
 		ELSE w=1;
 	RUN;
 
 When requesting a MCA on a preexisting contingency table, you must
 specify the number of variables that have been use to generate the
 table.
 
 Only one graph could be generated at a time. To browse the different
 factorial plans, recall the procedure and change the parameters.
 To obtain graphs for dimensions higher than 2, first change the
 parameter to set the number of working dimensions at a highler level.
 

 +------------------------------------------------------+   
 
 3 Known bugs / Limitations
 
 There is no automatic computation of a suitable maximum number of
 dimensions. Thus, the user may ask for more dimensions that possible
 and generate an error while executing. This could be seen in the log.
 For the computation of the maximum number of dimensions:
 in Simple Correspondence, take (Min(nm1,nm2)-1) where nm1 is the
 number of modalities of the first variable. For  Multiple
 Correspondence Analysis, each variable comes with (nm-1) dimensions.
 
 By default, only 1-10 dimensions are proposed, but you can enter
 the number of dimensions you need.
 
 When doing a Correspondence Analysis on raw data and recalling the
 task, it appears some times that the variable that play the role
 "raw variables to be crossed (rows)" are not saved and must be
 specified again.

 +------------------------------------------------------+   

 4 Copyrights
 
 SAS Enterprise Guide 2 is Copyright by SAS Institute Inc.
 
 This addon is copyleft, so that you can freely distribute it.
 The only condition is that you should provide this file
 with the DLL.
 
 +------------------------------------------------------+   
 
 5 Contact the author
 
 
 For any bug or suggestion, you can contact the author: 
 
 
 Eric Lecoutre <lecoutre@stat.ucl.ac.be>
 
 Institut de statistique, UCL
 Voie du Roman Pays, 20
 1348 Louvain-la-Neuve
 BELGIUM
 
 http://www.stat.ucl.ac.be