Re: Multi Correspondence Analysis
Posted by
Alex Reutter on
Aug 14, 2013; 3:35am
URL: http://spssx-discussion.165.s1.nabble.com/Multi-Correspondence-Analysis-tp5721553p5721572.html
I think there's a little confusion here.
* There is a CORRESPONDENCE procedure
whose syntax is pasted from the dialog accessed at Analyze
> Dimension Reduction > Correspondence Analysis...
* There is a MULTIPLE CORRESPONDENCE
procedure whose syntax is pasted from the dialog accessed at Analyze
> Dimension Reduction > Optimal Scaling...
Alex
From:
Ian Martin <[hidden email]>
To:
[hidden email],
Date:
08/13/2013 07:06 PM
Subject:
Re: Multi Correspondence
Analysis
Sent by:
"SPSSX(r)
Discussion" <[hidden email]>
Jon,
Oh, OK. This inspired me to RTFM, where I found that CA (as I would
use it, for a matrix of species X samples) is not under Dimension Reduction>Correspondence
Analysis, but rather under Dimension Reduction>Optimal Scaling>Multiple
Correspondence Analysis.
Thanks for putting me onto that, though I do think it's just a teeny bit
opaque to hide CA somewhere other than errr…. CA, just because it's _multiple_
CA. Especially when the syntax starts by calling the same procedure?
regards,
Ian
On Aug 13, 2013, at 6:21 PM, Jon K Peck <[hidden email]> wrote:
> Here is a response from the developer of the SPSS procedure.
>
> I'm on vacation, but I had quick look at it. First of all, SPSS CA
is also per Greenacre 1984, thus results should be the same as with Ade-4.
I searched the web for output of Ade-4 and found a document at
> ftp://pbil.univ-lyon1.fr/pub/mac/ADE/ADE4/DocThemPDF/Thema23.pdf
> containing a data example (data and syntax below). Results I see there
are equal to SPSS. What is called eigenvalues in Ade-4 is inertia in SPSS.
"Absolute contributions" in Ade-4 is "Contribution of points
to inertia of dimension" in SPSS in the Overview Row/Column points
table. What is called "factor" in Ade-4 is "Dimension"
in SPSS.
> The plots do not look the same, but it is not clear to me at a quick
glance what exactly is plotted in the document. It could be just a matter
of different normalization, and/or standardization, and/or measure.
> I hope this is helpful, if not then it would be useful to send data,
output and specific questions about the output
>
> data list free/rowid c7 TO c12 c1 TO c6 rowtot.
> begin data.
> 1 0 12 44 44 48 69 179 500 147 4 0 0 1047
> 2 4 4 30 16 20 11 17 75 70 1 0 0 248
> 3 0 7 9 13 7 14 18 74 16 0 0 0 158
> 4 1 0 0 3 5 6 11 98 6 0 0 0 130
> 5 0 4 38 18 21 9 16 87 218 51 2 0 464
> 6 0 1 2 5 3 7 1 11 14 2 0 0 46
> 7 1 0 3 2 0 7 5 1 0 1 0 0 20
> 8 0 7 28 17 9 4 9 9 10 2 0 0 95
> 9 0 4 8 12 2 3 5 5 1 0 0 1 41
> 10 1 12 20 13 9 2 0 0 2 1 0 0 60
> 11 1 43 31 7 2 1 1 1 4 5 0 1 97
> 12 4 68 53 15 3 0 0 0 0 5 25 7 180
> 13 0 14 3 0 0 0 0 0 0 0 7 1 25
> 14 3 184 105 34 5 0 0 0 0 11 83 13 438
> end data.
> CORRESPONDENCE
> TABLE = all (14,12)
> /DIMENSIONS = 2
> /MEASURE = CHISQ
> /STANDARDIZE = RCMEAN
> /NORMALIZATION = SYMMETRICAL
> /PRINT = TABLE RPOINTS CPOINTS rprof cprof
> /PLOT = NDIM(1,MAX) BIPLOT(20) rpoints cpoints.
>
>
> Jon Peck (no "h") aka Kim
> Senior Software Engineer, IBM
> [hidden email]
> phone: 720-342-5621
>
>
>
>
> From: Ian Martin <[hidden email]>
> To: Jon K Peck/Chicago/IBM@IBMUS, "[hidden email]"
<[hidden email]>,
> Date: 08/13/2013 01:16 PM
> Subject: Re: Multi Correspondence Analysis
>
>
>
> Jon,
>
> I've looked at the CA as implemented in SPSS, and can't see how it
equates to the CA I use frequently, which is in ADE-4 under R. The
CA I use is as per Greenacre 1984
> Greenacre, M. J. (1984) Theory and applications of correspondence
analysis, Academic Press, London.
> ADE-4:
> <http://link.springer.com/article/10.1023%2FA%3A1018513530268>
>
> cheers,
> Ian
> Ian D. Martin, Ph.D.
> Environmental Analysis & Consulting
>
> On Aug 13, 2013, at 12:40 PM, Jon K Peck <[hidden email]> wrote:
>
> > Multiple correspondence analysis is available in the Categories
option (Analyze > Dimension Reduction > Correspondence Analysis).
> >
> > You can read the help for this online in the IBM Info Center
at
> > http://pic.dhe.ibm.com/infocenter/spssstat/v21r0m0/index.jsp
> > under the Categories option.
> >
> > Jon Peck (no "h") aka Kim
> > Senior Software Engineer, IBM
> > [hidden email]
> > phone: 720-342-5621
> >
> >
> >
> >
> > From: Vik Rubenfeld <[hidden email]>
> > To: [hidden email],
> > Date: 08/13/2013 10:27 AM
> > Subject: Re: [SPSSX-L] Multi Correspondence
Analysis
> > Sent by: "SPSSX(r) Discussion"
<[hidden email]>
> >
> >
> >
> > Last time I looked into it, I did not find a way to do correspondence
analysis with SPSS. However that was a few years ago. I would love to find
a way to do CA in SPSS and I look forward to following this thread to find
out if there is one at this time.
> >
> > However, you can do CA using R Project, using an R Project package
such as "CA". http://www.r-project.org/
> >
> > Best,
> >
> >
> > -Vik
> >
> > On Aug 13, 2013, at 4:31 AM, mils <[hidden email]>
wrote:
> >
> > Hi everyone,
> >
> > I would like to run a Multi Correspondence Analysis (MCA). I
used to run MCA
> > with xlstat, so I am a bit confused on how to use it in SPSS.
I've attached
> > a file with the raw data and variables description.
> >
> > Do I have to transform the data?
> >
> > Any suggestions will be really appreciate it.
> >
> > Thanks in advance,
> >
> > Mils
> >
> > Multi_Correspondence_Analysis.xlsx
> > <http://spssx-discussion.1045642.n5.nabble.com/file/n5721553/Multi_Correspondence_Analysis.xlsx>
> >
> >
> >
> >
> > -----
> > mils
> > --
> > View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Multi-Correspondence-Analysis-tp5721553.html
> > Sent from the SPSSX Discussion mailing list archive at Nabble.com.
> >
> > =====================
> > To manage your subscription to SPSSX-L, send a message to
> > [hidden email] (not to SPSSX-L), with no body text
except the
> > command. To leave the list, send the command
> > SIGNOFF SPSSX-L
> > For a list of commands to manage subscriptions, send the command
> > INFO REFCARD
> >
>
>
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD