MVA Error in SPSS 22

[JosephYoungblood]

Wondering if anyone has had success with the work around for the following
error in v.22
MVA PROCEDURE ISSUES INSUFFICIENT MEMORY ERROR WITH SMALL ANALYSIS WHEN LONG
STRING VARIABLES ARE PRESENT IN FILE MVA

IBM/SPSS is aware of the problem, and recommended the following;
"For the time being, you can work around the issue either by removing string
variables with very long widths or using ALTER TYPE to trim off
trailing blanks. Open a syntax window and run the command:

ALTER TYPE ALL (A=AMIN).

Unfortuinately this doesn't solve the problem.

Thank you,

Joseph A. Youngblood

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

[Bruce Weaver]

The other option listed there is to remove the long string variables.  Presumably they are not needed by MVA, so remove them via /KEEP or /DROP sub-commands when opening the data file; or use DELETE VARIABLES to remove them after the file is open.

If you need those long string variables later for something else, re-open the file after running your analysis with MVA.

Joseph Youngblood wrote

Wondering if anyone has had success with the work around for the following
error in v.22
MVA PROCEDURE ISSUES INSUFFICIENT MEMORY ERROR WITH SMALL ANALYSIS WHEN LONG
STRING VARIABLES ARE PRESENT IN FILE MVA

IBM/SPSS is aware of the problem, and recommended the following;
"For the time being, you can work around the issue either by removing string
variables with very long widths or using ALTER TYPE to trim off
trailing blanks. Open a syntax window and run the command:

ALTER TYPE ALL (A=AMIN).

Unfortuinately this doesn't solve the problem.

Thank you,

Joseph A. Youngblood

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

[JosephYoungblood]

Thank you Bruce,

I'll add that to the growing list of things I have tried.

The caveat is that I can't /DELETE the string variables as they are integral
to MVA.  I ran the ALTER TYPE ALL (A=AMIN)to reduce string width, no avail.

BTW, have you heard anything about a pending SPSS 22 Fix Pack?

Joseph A. Youngblood

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

[David Marso]

Of course create some variable to merge later.
Assume active file.
Something like <UNTESTED>.
DATASET NAME activefile.
COMPUTE @mergevar=$CASENUM.
DATASET COPY longstrings.
DATASET ACTIVATE longstrings.
MATCH FILES / FILE * / KEEP @mergevar <list of long strings> .
DATASET ACTIVATE activefile.
DELETE VARIABLES <list of long strings> .

MVA blah blah blah ...

MATCH FILES / FILE * / FILE longstrings / BY @mergevar .

Bruce Weaver wrote

The other option listed there is to remove the long string variables.  Presumably they are not needed by MVA, so remove them via /KEEP or /DROP sub-commands when opening the data file; or use DELETE VARIABLES to remove them after the file is open.

If you need those long string variables later for something else, re-open the file after running your analysis with MVA.

Joseph Youngblood wrote

Wondering if anyone has had success with the work around for the following
error in v.22
MVA PROCEDURE ISSUES INSUFFICIENT MEMORY ERROR WITH SMALL ANALYSIS WHEN LONG
STRING VARIABLES ARE PRESENT IN FILE MVA

IBM/SPSS is aware of the problem, and recommended the following;
"For the time being, you can work around the issue either by removing string
variables with very long widths or using ALTER TYPE to trim off
trailing blanks. Open a syntax window and run the command:

ALTER TYPE ALL (A=AMIN).

Unfortuinately this doesn't solve the problem.

Thank you,

Joseph A. Youngblood

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

[David Marso]

I don't see how they would be integral to the MVA?
Please carefully explicate this.
You certainly aren't estimating missing values for them.
Maybe they are some sort of grouping variable?
In that case have you considered using AUTORECODE for them?
Really need context on this one.  I don't have the Missing Values Option, so really can't help you there aside from suggesting to apply common sense to the situation.
--

Joseph Youngblood wrote

Thank you Bruce,

I'll add that to the growing list of things I have tried.

The caveat is that I can't /DELETE the string variables as they are integral
to MVA.  I ran the ALTER TYPE ALL (A=AMIN)to reduce string width, no avail.

BTW, have you heard anything about a pending SPSS 22 Fix Pack?

Joseph A. Youngblood

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

[Bruce Weaver]

From the FM for MVA > VARIABLES Subcommand:

"String variables that are specified in the variable list, whether short or long, are automatically
defined as categorical. For a long string variable, only the first eight characters of the values
are used to distinguish categories."

So yes, AUTORECODE seems like the way to go.  I.e., AUTORECODE your string variables, then delete the original string variables from the file, and use the new AUTORECODEd variables in your MVA command.

David Marso wrote

I don't see how they would be integral to the MVA?
Please carefully explicate this.
You certainly aren't estimating missing values for them.
Maybe they are some sort of grouping variable?
In that case have you considered using AUTORECODE for them?
Really need context on this one.  I don't have the Missing Values Option, so really can't help you there aside from suggesting to apply common sense to the situation.
--

Joseph Youngblood wrote

Thank you Bruce,

I'll add that to the growing list of things I have tried.

The caveat is that I can't /DELETE the string variables as they are integral
to MVA.  I ran the ALTER TYPE ALL (A=AMIN)to reduce string width, no avail.

BTW, have you heard anything about a pending SPSS 22 Fix Pack?

Joseph A. Youngblood

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

[JosephYoungblood]

Thank you to all who contributed to my post, I did, and I hope others too
benefited from this inherently collaborative cross-fertilization forum.

Cheers,

Joseph A. Youngblood

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