topic Re: Multiple Correspondence Analysis - Enterprise miner (MCA) in SAS Data Science

Multiple Correspondence Analysis - Enterprise miner (MCA)

gabras — Tue, 17 Oct 2017 13:06:20 GMT

Hi, do you know how can i run MCA in Enterprise Miner?

I thought it was simple to find this but i can't.

Please help me.

Re: Multiple Correspondence Analysis - Enterprise miner (MCA)

DougWielenga — Fri, 20 Oct 2017 19:06:37 GMT

Multiple correspondence analysis (MCA) is a data analysis technique for nominal categorical data used to detect and represent underlying structures in a data set by representing data as points in a low-dimensional Euclidean space. SAS Enterprise Miner is designed for data mining (extremely large) data sets for which certain analytical methods like Multiple Correspondence Analysis would be unlikely to be used since plotting extremely large numbers of points is usually not useful.

There is an example of doing MCA in the CORRESP procedure documentation (part of SAS/STAT) which is available at

http://support.sas.com/documentation/cdl/en/statug/63962/HTML/default/viewer.htm#statug_corresp_sect026.htm

Although you could run code for the CORRESP procedure in SAS Enterprise Miner, there is no specific SAS Enterprise Miner node designed to do the same to my knowledge.

Hope this helps!

Doug

Re: Multiple Correspondence Analysis - Enterprise miner (MCA)

gabras — Sat, 21 Oct 2017 06:33:57 GMT

Hi Doug, thank you for the answer.
Since Enterprise Miner is built for data mining, would you suggest any orher techniques which could i use instead of MCA in order to find associations between nominal variables and factors from the reduction of theese to use in further analysis?

It would be very helpful!

Thanks again

Re: Multiple Correspondence Analysis - Enterprise miner (MCA)

DougWielenga — Mon, 23 Oct 2017 13:43:32 GMT

There are several techniques for data reduction and interpretation but the choice among which ones to try depends greatly on the nature of your data (number and type of fields, number of observations, presence of and/or type of target/outcome variable of interest) as well as the business problem you are trying to solve. You must also consider the available computing power you have to solve the problem since some approaches might be more computationally intensive than others. It would be helpful if you could describe the nature of your data as well as the objective you hope to accomplish after analyzing your data.

Cordially,

Doug