Seeds for Quick Cluster - Do They Nullify the Order of the Records Going into the Procedure? - I do not think so

[Feinstein, Zachary]

One more question while we are on the topic.  For the below query I
found out I had too many cases/records for PROXIMITIES to work in V15.

But let us say it did work as it does in V16.  If I use the results from
the Cluster Analysis, after the PROXIMITIES command, and use those as
seeds for QUICK CLUSTER will the order of my records be a null issue
using seeds?

For example- given I was not able to do the PROXIMITIES command I went
straight to QUICK CLUSTER.

- I first sorted my data in ascending order of my id then ran it.
- I then sorted my data in descending order of my id then ran it again
too.
- I then crosstabulated the results as an index of sorts of the
stability of the solution.

If I feed seeds into the various sortings does the sorting still affect
the results?  I think they do.  I would imagine the seeds just nudge the
results toward a more stable solution but it still is heavily dependent
on record order for the QUICK CLUSTER procedure.  I think too that if I
specify some subcommand (such as NO UPDATE) then it will just lump the
records in with the original solution without needing to worry about
record order.

Thank you.

Zachary

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
ViAnn Beadle
Sent: Friday, November 02, 2007 4:49 PM
To: [hidden email]
Subject: Re: Trouble Now With PROXIMITIES for Cluster Analyses

On the face of it this doesn't sound like you ran out of memory--this
error message means that SPSS couldn't read something from memory which
could mean almost anything. I'd ask SPSS technical support about this.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Feinstein, Zachary
Sent: Friday, November 02, 2007 2:20 PM
To: [hidden email]
Subject: Trouble Now With PROXIMITIES for Cluster Analyses

Okay- now I am onto a problem that I am having with the cluster analysis
phase of this project of mine.  Thank you to everyone who responded
before to help me kickstart where I screwed up with redundant variables.

I wish to do the hierarchical cluster analysis for some of the variables
from my Canonical Correlations.  I have almost 2,500 respondents and I
get rid of all the extra variables with the following:

MATCH FILES
 /FILE = *
 /KEEP = id zcv1 zcv2 zcv3 zcv4.
EXECUTE.

Then I do PROXIMITIES with the following:

PROXIMITIES zcv1 TO zcv4
 /MATRIX OUT ('C:\temp2\spssclus.tmp')
 /VIEW = CASE
 /MEASURE = SEUCLID
 /PRINT = NONE
 /STANDARDIZE = NONE.

But it very consistently shuts SPSS down first saying "The Processor has
terminated due to an unrecoverable error at 0x0abf9ecd fetching from
0x07a11000."

I am using SPSS 15.0.1.

I know back in the old days I would have problems with the CLUSTER
procedure next- I would set MXMEMORY to something higher before the
CLUSTER procedure then set it back to something low after that.  I
recall sometime within the past few versions that SPSS has said one
never needs to vary the memory anymore.  An example from before follows:

SET MXMEMORY = 56000.
EXECUTE.

CLUSTER
 /MATRIX IN ('C:\WINDOWS\TEMP\spssclus.tmp')  /METHOD WARD  /PRINT
SCHEDULE CLUSTER(3, 10)  /PRINT DISTANCE  /PLOT NONE  /SAVE CLUSTER(3,
10).

SET MXMEMORY = 14000.
EXECUTE.

CACHE.
EXECUTE.

But I cannot even get to that because of the PROXIMITIES shutdown.  By
the way- After I do this, eventually, I use the results as SEEDS for
multiple QUICK CLUSTER runs.

Any thoughts are very much appreciated.

Zachary

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Peck, Jon
Sent: Friday, November 02, 2007 2:36 PM
To: [hidden email]
Subject: Re: Trouble Isolating Duplicate Variables in MANOVA Command

I have built a little Python function called dupVarnameCheck to expand a
variable list against the dictionary of the active dataset and check for
duplicate entries.  I will be adding this to the spssaux2 module on SPSS
Developer Central, probably later today.

Regards,
Jon Peck



-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
James C. Whanger
Sent: Friday, November 02, 2007 1:11 PM
To: [hidden email]
Subject: Re: [SPSSX-L] Trouble Isolating Duplicate Variables in MANOVA
Command

You're quite welcome.  It was just a sharp eye.  Unfortunately, I was
not able to come up with an automated solution.  Let me know if you
discover one.


On Nov 2, 2007 2:15 PM, Feinstein, Zachary
<[hidden email]> wrote:

> Was that a sharp eye or did you use some kind of automated process?
>
> Thank you very much.
>
> Zachary
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf
> Of James C. Whanger
> Sent: Friday, November 02, 2007 11:25 AM
> To: [hidden email]
>
> Subject: Re: Trouble Isolating Duplicate Variables in MANOVA Command
>
> My apologies.  I responded to the wrong posting the first time.
>
> Zachary,
>
> The variable "zq721a1m4" is used twice.  It appears the second time it

> was intended to be changed to "zq721a2m4".
>
> Good luck with your analyses,
>
> Jim
>
> On Nov 2, 2007 11:26 AM, Feinstein, Zachary
> <[hidden email]> wrote:
> > I have about 200 variables that I am submitting to a MANOVA to see
> > if any of the variables are linearly dependent before I run a
> > Canonical Correlation.  The message I get when I try to run it is:
> >
> > >Error # 12160
> >
> > >Duplicate factor or variable name found on MANOVA command.
> >
> > >This command not executed.
> >
> > I have so many variables in there that it is difficult for me to
> > isolate where the problem is.  Now if run Frequencies on all of the
> > variables then it will always run even if there are duplicates, so
> > that does not help me.  It does not burp and say where the duplicate
> is.
> >
> > I also exported the fieldnames to Excel thinking it would list out
> > the
>
> > various duplicate names, I sorted on the fieldnames once I
> > transposed the Excel file, but I found no duplicates.  I wonder if
> > SAVE TRANSLATE
>
> > OUTFILE is similar to Frequencies where it will list the variable
> > only
>
> > once?
> >
> > Can anyone suggest a way to isolate what the duplicate variable
> > names are?  Thank you in advance.
> >
> > Zachary
> >
> > [hidden email]
> >
> > zpwin18_1 TO zpwin18_9
> >
> > zpwin48_1 TO zpwin48_5
> >
> > zpwin78_1 TO zpwin78_4
> >
> > zpwin08_1 TO zpwin08_5
> >
> > zq660
> >
> > zq1201a1 TO zq1201a23
> >
> > zq1300
> >
> > zq1305
> >
> > zq1310
> >
> > zq1315
> >
> > zq1320
> >
> > zq1340
> >
> > zq1345
> >
> > zq1346
> >
> > zq1385 zq1386 zq1387
> >
> > WITH
> >
> > zq445_1 zq445_2
> >
> > zq102
> >
> > zq105
> >
> > zq615_1 zq615_2 zq615_3 zq615_5 zq615_6
> >
> > zq450_1 zq450_2
> >
> > zq455
> >
> > zq600_1 zq600_2 zq600_3 zq600_4 zq600_5
> >
> > zq452_1 zq452_2 zq452_3 zq452_4 zq452_5 zq452_9
> >
> > zq605m1 TO zq605m99
> >
> > zq630
> >
> > zq650_1 zq650_2
> >
> > zq655_1 zq655_2
> >
> > zq665_1 zq665_2 zq665_3 zq665_6 zq665_7 zq665_9
> >
> > zq701a1 TO zq701a5
> >
> > zq706a1 TO zq706a7
> >
> > zq715
> >
> > zq721a1m2 zq721a1m3 zq721a1m4 zq721a1m9 zq721a1m10 zq721a1m11
> > zq721a1m12
> > zq721a1m14 zq721a1m15 zq721a1m16
> >
> > zq721a2m2 zq721a2m3 zq721a1m4 zq721a2m9 zq721a2m10 zq721a2m11
> > zq721a2m12
> > zq721a2m15 zq721a2m16
> >
> > zq732a2_3 zq732a3_3 zq732a3_4 zq732a4_3 zq732a4_4 zq732a9_3
> > zq732a9_4
> > zq732a10_3 zq732a10_4 zq732a11_3 zq732a11_4 zq732a12_3 zq732a12_4
> > zq732a15_3
> >
> > zq735a9_3 zq735a9_4 zq735a10_3 zq735a11_3 zq735a11_4 zq735a12_3
> > zq735a12_4
> >
> > zq725
> >
> > zq741a1 TO zq741a12
> >
> > zq202_1 zq202_2 zq202_3 zq202_6
> >
> > zq204 zq206
> >
> > zq216_3 zq216_4 zq216_5 zq216_6 zq216_7 zq216_70
> >
> > zq232
> >
> >
> >
> > 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
>

=====================
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

=====================
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

=====================
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