SPSS 22.0.0.1 on XUbuntu 12.4 crashes on cluster analyses

[Andreas Sprenger]

Hi,

working for years now with SPSS on DOS and Windows machines I have to work now on Linux systems. I can reproduce crashes on different hardware with the same software (XUbuntu, SPSS 22) and different datasets using cluster analyses. This does not happen on Windows systems. I should try it on Mac ...

Imagine a dataset with twelve variables x1 .. x12. Running proximities works fine, running cluster subsequently induces a crash in the java virtual machine. The users gets a message saying something like "connection to remote server failed". Any ideas?

Andreas

--
Here's the syntax:

DATASET DECLARE D0.14383517414668257.
PROXIMITIES   x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12
  /MATRIX OUT(D0.14383517414668257)
  /VIEW=CASE
  /MEASURE=SEUCLID
  /PRINT NONE
  /STANDARDIZE=VARIABLE Z.
 
CLUSTER
  /MATRIX IN(D0.8205591979286913)
  /METHOD BAVERAGE
  /PRINT SCHEDULE
  /PLOT DENDROGRAM.
Dataset Close D0.8205591979286913.
 

[Art Kendall]

The size of the proximities matrix is 8 bytes * the number of cases * the number of cases.

how many cases are in DATASET  D0.14383517414668257?

What happens if you use a shorter dataset name?

On 4/1/2014 9:37 AM, Andreas Sprenger [via SPSSX Discussion] wrote:

Hi,

working for years now with SPSS on DOS and Windows machines I have to work now on Linux systems. I can reproduce crashes on different hardware with the same software (XUbuntu, SPSS 22) and different datasets using cluster analyses. This does not happen on Windows systems. I should try it on Mac ...

Imagine a dataset with twelve variables x1 .. x12. Running proximities works fine, running cluster subsequently induces a crash in the java virtual machine. The users gets a message saying something like "connection to remote server failed". Any ideas?

Andreas

--
Here's the syntax:

 
PROXIMITIES   x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12
  /MATRIX OUT(D0.14383517414668257)
  /VIEW=CASE
  /MEASURE=SEUCLID
  /PRINT NONE
  /STANDARDIZE=VARIABLE Z.
 
CLUSTER
  /MATRIX IN(D0.8205591979286913)
  /METHOD BAVERAGE
  /PRINT SCHEDULE
  /PLOT DENDROGRAM.
Dataset Close D0.8205591979286913.
 

---

If you reply to this email, your message will be added to the discussion below:

http://spssx-discussion.1045642.n5.nabble.com/SPSS-22-0-0-1-on-XUbuntu-12-4-crashes-on-cluster-analyses-tp5725146.html

To start a new topic under SPSSX Discussion, email [hidden email]
To unsubscribe from SPSSX Discussion, click here.
NAML

[Moon Kid]

Just by side: You are running SPSS 22 on XUbuntu 12.04?

How? I was never able to do that on Kubuntu 13.10. And the IBM support
fucks me up and tell me they support only three special Linux-Distris
(RedHat and Debian) for there "plattformindependent" nice Java-written
SPSS.

Maybe you know this error (doesn't matter if root or not)?
[bash]
user@kubuntu-monster:~/share$ sudo ./SPSS_Statistics_22_TR_Lx_.bin
[sudo] password for user:
Preparing to install...
Extracting the JRE from the installer archive...
Unpacking the JRE...
Extracting the installation resources from the installer archive...
Configuring the installer for this system's environment...

Launching installer...

/tmp/install.dir.8824/Linux/resource/jre/jre/bin/java:
2: /tmp/install.dir.8824/Linux/resource/jre/jre/bin/java: Syntax error: "(" unexpected
[/bash]

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

[Andreas Sprenger]

Hi,

N = 30

The filename (D0.14383517414668257) is generated by SPSS. I will try it with a shorter filename but I think it won't help...

[Andreas Sprenger]

Moon Kid wrote

Just by side: You are running SPSS 22 on XUbuntu 12.04?

Correct. Standard installation with xfce 4.2. No problem except cluster analyses...

Moon Kid wrote

Maybe you know this error (doesn't matter if root or not)?
[bash]
user@kubuntu-monster:~/share$ sudo ./SPSS_Statistics_22_TR_Lx_.bin
[sudo] password for user:
Preparing to install...
Extracting the JRE from the installer archive...
Unpacking the JRE...
Extracting the installation resources from the installer archive...
Configuring the installer for this system's environment...

Launching installer...

/tmp/install.dir.8824/Linux/resource/jre/jre/bin/java:
2: /tmp/install.dir.8824/Linux/resource/jre/jre/bin/java: Syntax error: "(" unexpected
[/bash]

No. Installation proceeded without errors, FixPack1 too.

Try downloading current (trial) version from IBM. Maybe the syntax error has been fixed...

Andreas

[Art Kendall]

Are you sure that works on any platform.

You have the name of the output dataset from PROXIMITIES is not the same as the input dataset name in CLUSTER.

try the following syntax on your different platforms.
Please let us know what you find.

new file.
input program.
   vector x (12,f3).
   loop id = 1 to 30.
      loop #p = 1 to 12.
         compute x(#p) = rnd(rv.normal(50,10)).
      end loop.
      end case.
   end loop.
   end file.
end input program.
DATASET DECLARE MyProx.
PROXIMITIES   x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12
  /MATRIX OUT(MyProx)
  /VIEW=CASE
  /MEASURE=SEUCLID
  /PRINT NONE
  /STANDARDIZE=VARIABLE Z.
 
CLUSTER
  /MATRIX IN(MyProx)
  /METHOD BAVERAGE
  /PRINT SCHEDULE
  /PLOT DENDROGRAM.
Dataset Close MyProx.
On 4/1/2014 9:37 AM, Andreas Sprenger [via SPSSX Discussion] wrote:

Hi,

working for years now with SPSS on DOS and Windows machines I have to work now on Linux systems. I can reproduce crashes on different hardware with the same software (XUbuntu, SPSS 22) and different datasets using cluster analyses. This does not happen on Windows systems. I should try it on Mac ...

Imagine a dataset with twelve variables x1 .. x12. Running proximities works fine, running cluster subsequently induces a crash in the java virtual machine. The users gets a message saying something like "connection to remote server failed". Any ideas?

Andreas

--
Here's the syntax:

DATASET DECLARE D0.14383517414668257.
PROXIMITIES   x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12
  /MATRIX OUT(D0.14383517414668257)
  /VIEW=CASE
  /MEASURE=SEUCLID
  /PRINT NONE
  /STANDARDIZE=VARIABLE Z.
 
CLUSTER
  /MATRIX IN(D0.8205591979286913)
  /METHOD BAVERAGE
  /PRINT SCHEDULE
  /PLOT DENDROGRAM.
Dataset Close D0.8205591979286913.
 

---

If you reply to this email, your message will be added to the discussion below:

http://spssx-discussion.1045642.n5.nabble.com/SPSS-22-0-0-1-on-XUbuntu-12-4-crashes-on-cluster-analyses-tp5725146.html

To start a new topic under SPSSX Discussion, email [hidden email]
To unsubscribe from SPSSX Discussion, click here.
NAML

[Andreas Sprenger]

Thanks for the help. You're right, filenames were not the same. Please keep in mind that the syntax was generated by SPSS (on Linux). Correcting the filenames helps on Windows machines but not on Linux.

In order to play safe I tried your syntax which works well on Windows machines but crashes SPSS on Linux systems. So I think it is a bug in the Linux version. I will now install SPSS on a Mac in order to see whether it crashes as well.

Stay tuned.

Andreas

[Andreas Sprenger]

Update on SPSS on XUbuntu and cluster analyses

Meanwhile we are working on XUbuntu 14.04 LTS: SPSS runs more stable which means one crash in 90 minutes compared to one crash in 15 minutes before.

There are two bugs in the cluster analysis on Linux systems (not on Windows or Mac), that result in an "unknown error" which stops the SPSS processor.

1) The usage of a string variable as a case label induces a crash (subcommand /ID = varname)

2) If you want to have z-scores within the cluster analysis, the same error occurs. If you want to use z-transformed variables create them before by descriptives

Details to bug 2)
If you request z-scores in the CLUSTER menu SPSS creates a syntax which additionally uses PROXIMITIES (as on SPSSX in the 80ies). Proximities are written into a temporary dataset without a problem. The CLUSTER command with a subcommand /MATRIX IN(Datasetname) induces the crash.

Is there no bug testing at SPSS?

Cheers

Andreas