Longitudinal data / data processing

[Jennifer Thompson]

Dear SPSSers,

I wonder if anyone could suggest a solution to this data processing /
manipulation problem?

*Background and description of data*
I am working on a longitudinal data set that consists of psychiatric
symptoms, coded on a 5-point (0-4) scale.  The data set is 'long' with a
separate case for each observation.  Participants were interviewed annually
and the number of observations per participant is variable.  Each
participant has a unique ID code (combination of letters and digits) and a
date is recorded for each interview.

*Question*
I would like to know how many participants have had a score of 2 or more
(this being the accepted criterion for the symptom being present) for each
symptom at _any_ of the interviews during the follow up period.

So, I guess I want to create a new variable (e.g. 'ever_panic') that is
coded '1' if a given symptom (e.g. 'panic') is scored 2 or higher for _any_
of the interviews with an individual participant.  Or alternatively, to
generate an output that lists the ID codes and whether or not the symptom
was present at any time?

I can only think of tortuously long-winded ways to do this, which is not
ideal because the process will need to be repeated for 30+ items!  Can any
suggest a solution?  I am using SPSS 14.

I would be very grateful for any suggestions!  Thanks for reading.

Best wishes,

Jennifer

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

[Sara House]

Jennifer -

I'm not directly answering your question (sorry) but I do have a resource to point you toward that may be useful.  I recently attended a workshop at APS by Don Hedeker in which he discussed longitudinal analysis in SPSS.  Here is his webpage:

http://tigger.uic.edu/~hedeker/

All of the handouts, datasets and syntax files are available for download.  He also gives SAS code if you would prefer to use SAS for your analyses.  Hope this helps.

Sara

Sara M. House, M.A.
Adjunct Faculty
Loyola University Chicago, Psychology Department
Email: [hidden email]
Teaching: Research Methods, Psychology & Law

AND

Data Analyst
Chicago Public Schools, Department of Program Evaluation
Email: [hidden email]

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

[Burleson,Joseph A.]

Jennifer:

Try the COUNT transform command.

COUNT NEWVAR_SYMA=syma1 syma2 syma3 (2 THRU 5).
COUNT NEWVAR_SYMB=SYMb1 SYMb2 SYMb3 (2).
EXECUTE.

For the above, symptoms are represented by letters, the annual variables
are represented by numbers.

For each symptom, do the above, with the listing of the repeated
variable/year after the equal sign. In parentheses are the values (2
through 5, in your case) for which you wish a "running count" to be
created, for each symptom, for each subject. Therefore, if a given
subject has ten years of data, and eight times s/he had a value of 2 or
more, then NEWVAR = 8, for that subject.

Then do a FREQUENCIES for each NEWVAR_SYMx.

See if that works for you.

Joe Burleson

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Jennifer Thompson
Sent: Thursday, May 29, 2008 12:46 PM
To: [hidden email]
Subject: Longitudinal data / data processing

Dear SPSSers,

I wonder if anyone could suggest a solution to this data processing /
manipulation problem?

*Background and description of data*
I am working on a longitudinal data set that consists of psychiatric
symptoms, coded on a 5-point (0-4) scale.  The data set is 'long' with a
separate case for each observation.  Participants were interviewed
annually
and the number of observations per participant is variable.  Each
participant has a unique ID code (combination of letters and digits) and
a
date is recorded for each interview.

*Question*
I would like to know how many participants have had a score of 2 or more
(this being the accepted criterion for the symptom being present) for
each
symptom at _any_ of the interviews during the follow up period.

So, I guess I want to create a new variable (e.g. 'ever_panic') that is
coded '1' if a given symptom (e.g. 'panic') is scored 2 or higher for
_any_
of the interviews with an individual participant.  Or alternatively, to
generate an output that lists the ID codes and whether or not the
symptom
was present at any time?

I can only think of tortuously long-winded ways to do this, which is not
ideal because the process will need to be repeated for 30+ items!  Can
any
suggest a solution?  I am using SPSS 14.

I would be very grateful for any suggestions!  Thanks for reading.

Best wishes,

Jennifer

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

[Jennifer Thompson]

Hi Joe,

Many thanks for your response.  Unless I've misunderstood you, I think the
method you suggest would work perfectly if my data were 'wide', with all the
observations for each individual in one row /case, but each observation is
actually a seperate row/case.  I probably did not explain this clearly!  See
(fabricated) example below.

Best wishes,

Jennifer

id     date        syma    symb   symc  ----etc
jt76   01.05.98   0          1        0
jt76   01.05.99   1          2        0
jt76   01.05.00   2          2        3
jt76   01.05.01   0          0        0
lj75   01.06.98   2          3        2
lj75   01.05.99   2          4        2
lj75   02.05.00   2          2        3
lj75   01.05.01   1          2        3
ah71 01.05.97   1          0        0
ah71 01.05.98    0          0        0
ah71 01.05.99    0          0        0
ah71 01.05.00    1          1        0
sm68  01.05.00   1          2        0
sm68 01.05.01    0          0        0sm68 01.05.02    1          1        1


Jennifer:

>
> Try the COUNT transform command.
>
> COUNT NEWVAR_SYMA=syma1 syma2 syma3 (2 THRU 5).
> COUNT NEWVAR_SYMB=SYMb1 SYMb2 SYMb3 (2).
> EXECUTE.
>
> For the above, symptoms are represented by letters, the annual variables
> are represented by numbers.
>
> For each symptom, do the above, with the listing of the repeated
> variable/year after the equal sign. In parentheses are the values (2
> through 5, in your case) for which you wish a "running count" to be
> created, for each symptom, for each subject. Therefore, if a given
> subject has ten years of data, and eight times s/he had a value of 2 or
> more, then NEWVAR = 8, for that subject.
>
> Then do a FREQUENCIES for each NEWVAR_SYMx.
>
> See if that works for you.
>
> Joe Burleson
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
> Jennifer Thompson
> Sent: Thursday, May 29, 2008 12:46 PM
> To: [hidden email]
> Subject: Longitudinal data / data processing
>
> Dear SPSSers,
>
> I wonder if anyone could suggest a solution to this data processing /
> manipulation problem?
>
> *Background and description of data*
> I am working on a longitudinal data set that consists of psychiatric
> symptoms, coded on a 5-point (0-4) scale.  The data set is 'long' with a
> separate case for each observation.  Participants were interviewed
> annually
> and the number of observations per participant is variable.  Each
> participant has a unique ID code (combination of letters and digits) and
> a
> date is recorded for each interview.
>
> *Question*
> I would like to know how many participants have had a score of 2 or more
> (this being the accepted criterion for the symptom being present) for
> each
> symptom at _any_ of the interviews during the follow up period.
>
> So, I guess I want to create a new variable (e.g. 'ever_panic') that is
> coded '1' if a given symptom (e.g. 'panic') is scored 2 or higher for
> _any_
> of the interviews with an individual participant.  Or alternatively, to
> generate an output that lists the ID codes and whether or not the
> symptom
> was present at any time?
>
> I can only think of tortuously long-winded ways to do this, which is not
> ideal because the process will need to be repeated for 30+ items!  Can
> any
> suggest a solution?  I am using SPSS 14.
>
> I would be very grateful for any suggestions!  Thanks for reading.
>
> Best wishes,
>
> Jennifer
>
> =====================
> 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

[ViAnn Beadle]

If you're trying to select individuals for further analysis you can easily
do this by aggregating on ID code and using the PGT function which will
provide the percentage greater than 2 for your symptom by ID. You end up
with one case per ID or with MODE=ADDVARIABLES, that value attached to each
case.


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Jennifer Thompson
Sent: Thursday, May 29, 2008 10:46 AM
To: [hidden email]
Subject: Longitudinal data / data processing

Dear SPSSers,

I wonder if anyone could suggest a solution to this data processing /
manipulation problem?

*Background and description of data*
I am working on a longitudinal data set that consists of psychiatric
symptoms, coded on a 5-point (0-4) scale.  The data set is 'long' with a
separate case for each observation.  Participants were interviewed annually
and the number of observations per participant is variable.  Each
participant has a unique ID code (combination of letters and digits) and a
date is recorded for each interview.

*Question*
I would like to know how many participants have had a score of 2 or more
(this being the accepted criterion for the symptom being present) for each
symptom at _any_ of the interviews during the follow up period.

So, I guess I want to create a new variable (e.g. 'ever_panic') that is
coded '1' if a given symptom (e.g. 'panic') is scored 2 or higher for _any_
of the interviews with an individual participant.  Or alternatively, to
generate an output that lists the ID codes and whether or not the symptom
was present at any time?

I can only think of tortuously long-winded ways to do this, which is not
ideal because the process will need to be repeated for 30+ items!  Can any
suggest a solution?  I am using SPSS 14.

I would be very grateful for any suggestions!  Thanks for reading.

Best wishes,

Jennifer

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

[hillel vardi]

Shalom

you can use the fallowing code .

recode symptom1 to symptom10(0 1=0)(2 3 4=1) .
aggregate outfile=* / by  ID / symptom1 to symptom10=sum(symptom1 to
symptom10) .
recode symptom1 to symptom10(1 thru hi=1).

that will give you a file with a line for every Participant and  value 1
for each symptom if that symptom had value of  2 or higher in any of the
Participant observation.

you can then match that line to the original file if you need it .

Hillel Vardi
BGU



Jennifer Thompson wrote:

> Dear SPSSers,
>
> I wonder if anyone could suggest a solution to this data processing /
> manipulation problem?
>
> *Background and description of data*
> I am working on a longitudinal data set that consists of psychiatric
> symptoms, coded on a 5-point (0-4) scale.  The data set is 'long' with a
> separate case for each observation.  Participants were interviewed annually
> and the number of observations per participant is variable.  Each
> participant has a unique ID code (combination of letters and digits) and a
> date is recorded for each interview.
>
> *Question*
> I would like to know how many participants have had a score of 2 or more
> (this being the accepted criterion for the symptom being present) for each
> symptom at _any_ of the interviews during the follow up period.
>
> So, I guess I want to create a new variable (e.g. 'ever_panic') that is
> coded '1' if a given symptom (e.g. 'panic') is scored 2 or higher for _any_
> of the interviews with an individual participant.  Or alternatively, to
> generate an output that lists the ID codes and whether or not the symptom
> was present at any time?
>
> I can only think of tortuously long-winded ways to do this, which is not
> ideal because the process will need to be repeated for 30+ items!  Can any
> suggest a solution?  I am using SPSS 14.
>
> I would be very grateful for any suggestions!  Thanks for reading.
>
> Best wishes,
>
> Jennifer
>
> =====================
> 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

[Carsten Pauck]

Hello,
another idea is to restructure the dataset. As a result you have one
row for every patient (=id):
You start with the data:
   ID        DATE                SYMA    SYMB    SYMC
   1,00 01-MAY-1998     ,00    1,00     ,00
   1,00 01-MAY-1999    1,00    2,00     ,00
   1,00 01-MAY-2000    2,00    2,00    3,00
   2,00 02-MAY-1998     ,00     ,00     ,00
   2,00 02-MAY-1999    2,00    3,00    2,00
   3,00 03-MAY-1998    2,00    4,00    2,00
   3,00 03-MAY-1999    2,00    2,00    3,00
   3,00 03-MAY-2000    1,00    2,00    3,00
   3,00 03-MAY-2001    1,00     ,00     ,00
And you restructure the data: Data -> restructure ... which will generate
the
following syntax:
SORT CASES BY id .
CASESTOVARS
/ID = id
/GROUPBY = VARIABLE .
Which will transform the data in switching vars to cases.

syma.1  syma.2 syma.3 syma.4    symb.1  etc
   1,00 01-MAY-1998 01-MAY-1999 01-MAY-2000 .                      ,00
 1,00    2,00 .    1,00    2,00    2,00 .     ,00     ,00    3,00 .
   2,00 02-MAY-1998 02-MAY-1999 . .
,00    2,00 . .     ,00    3,00 . .     ,00    2,00 . .
   3,00 03-MAY-1998 03-MAY-1999 03-MAY-2000 03-MAY-2001    2,00    2,00
 1,00 1,00    4,00    2,00    2,00     ,00    2,00    3,00    3,00     ,00


Now it is possible to analyse the data in the way you are used to (count,
etc ..).
Carsten.


2008/5/29 Jennifer Thompson <[hidden email]>:
> Dear SPSSers,
>
> I wonder if anyone could suggest a solution to this data processing /
> manipulation problem?
>
> *Background and description of data*
> I am working on a longitudinal data set that consists of psychiatric
> symptoms, coded on a 5-point (0-4) scale.  The data set is 'long' with a
> separate case for each observation.  Participants were interviewed
annually

> and the number of observations per participant is variable.  Each
> participant has a unique ID code (combination of letters and digits) and a
> date is recorded for each interview.
>
> *Question*
> I would like to know how many participants have had a score of 2 or more
> (this being the accepted criterion for the symptom being present) for each
> symptom at _any_ of the interviews during the follow up period.
>
> So, I guess I want to create a new variable (e.g. 'ever_panic') that is
> coded '1' if a given symptom (e.g. 'panic') is scored 2 or higher for

_any_

> of the interviews with an individual participant.  Or alternatively, to
> generate an output that lists the ID codes and whether or not the symptom
> was present at any time?
>
> I can only think of tortuously long-winded ways to do this, which is not
> ideal because the process will need to be repeated for 30+ items!  Can any
> suggest a solution?  I am using SPSS 14.
>
> I would be very grateful for any suggestions!  Thanks for reading.
>
> Best wishes,
>
> Jennifer
>
> =====================
> 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

[Jennifer Thompson]

Many thanks to Joe, Carsten, Hillel & ViAnn for all the very helpful
responses to this query (I have included the original query and responses
below).  Restructuring the datafile from long/stacked to wide/flat would
seem to be best solution.

Best wishes,

Jennifer


From: *Jennifer Thompson* <[hidden email]>
Date: 2008/5/29
To: [hidden email]


Dear SPSSers,

I wonder if anyone could suggest a solution to this data processing /
manipulation problem?

*Background and description of data*
I am working on a longitudinal data set that consists of psychiatric
symptoms, coded on a 5-point (0-4) scale.  The data set is 'long' with a
separate case for each observation.  Participants were interviewed annually
and the number of observations per participant is variable.  Each
participant has a unique ID code (combination of letters and digits) and a
date is recorded for each interview.

*Question*
I would like to know how many participants have had a score of 2 or more
(this being the accepted criterion for the symptom being present) for each
symptom at _any_ of the interviews during the follow up period.

So, I guess I want to create a new variable (e.g. 'ever_panic') that is
coded '1' if a given symptom (e.g. 'panic') is scored 2 or higher for _any_
of the interviews with an individual participant.  Or alternatively, to
generate an output that lists the ID codes and whether or not the symptom
was present at any time?

I can only think of tortuously long-winded ways to do this, which is not
ideal because the process will need to be repeated for 30+ items!  Can any
suggest a solution?  I am using SPSS 14.

I would be very grateful for any suggestions!  Thanks for reading.

Best wishes,

Jennifer
----------
From: *Burleson,Joseph A.* <[hidden email]>
Date: 2008/5/29
To: Jennifer Thompson <[hidden email]>, [hidden email]


Jennifer:

Try the COUNT transform command.

COUNT NEWVAR_SYMA=syma1 syma2 syma3 (2 THRU 5).
COUNT NEWVAR_SYMB=SYMb1 SYMb2 SYMb3 (2).
EXECUTE.

For the above, symptoms are represented by letters, the annual variables
are represented by numbers.

For each symptom, do the above, with the listing of the repeated
variable/year after the equal sign. In parentheses are the values (2
through 5, in your case) for which you wish a "running count" to be
created, for each symptom, for each subject. Therefore, if a given
subject has ten years of data, and eight times s/he had a value of 2 or
more, then NEWVAR = 8, for that subject.

Then do a FREQUENCIES for each NEWVAR_SYMx.

See if that works for you.

Joe Burleson
=====================
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
----------
From: *Jennifer Thompson* <[hidden email]>
Date: 2008/5/29
To: "Burleson,Joseph A." <[hidden email]>
Cc: [hidden email]


Hi Joe,

Many thanks for your response.  Unless I've misunderstood you, I think the
method you suggest would work perfectly if my data were 'wide', with all the
observations for each individual in one row /case, but each observation is
actually a seperate row/case.  I probably did not explain this clearly!  See
(fabricated) example below.

Best wishes,

Jennifer

id     date        syma    symb   symc  ----etc
jt76   01.05.98   0          1        0
jt76   01.05.99   1          2        0
jt76   01.05.00   2          2        3
jt76   01.05.01   0          0        0
lj75   01.06.98   2          3        2
lj75   01.05.99   2          4        2
lj75   02.05.00   2          2        3
lj75   01.05.01   1          2        3
ah71 01.05.97   1          0        0
ah71 01.05.98    0          0        0
ah71 01.05.99    0          0        0
ah71 01.05.00    1          1        0
sm68  01.05.00   1          2        0
sm68 01.05.01    0          0        0 sm68 01.05.02    1          1
1

----------
From: *ViAnn Beadle* <[hidden email]>
Date: 2008/5/29
To: Jennifer Thompson <[hidden email]>, [hidden email]


If you're trying to select individuals for further analysis you can easily
do this by aggregating on ID code and using the PGT function which will
provide the percentage greater than 2 for your symptom by ID. You end up
with one case per ID or with MODE=ADDVARIABLES, that value attached to each
case.
----------
From: *Burleson,Joseph A.* <[hidden email]>
Date: 2008/5/29
To: Jennifer Thompson <[hidden email]>


 Jennifer:



No problem: Simply use the RESTRUCTURE command in SPSS (best to use the
"Wizard" that does it for you interactively, see the SPSS documentation).
That will make a "flat" file out of your "stacked" file. The wizard asks how
many types (30+ in your case) or groups of variables you want to
restructure. It asks what is the maximum number of "stacked" rows per case.
It also asks which variables are constant (e.g., gender), since it only has
to ask for them once. You can keep as many or as few variables in your new
flat file as you wish. ID, of course, has to be the "link" variable, so take
note of how this is done in the WIZARD procedures.



Then do the COUNT on the flat file.



Joe


 ------------------------------

*From:* Jennifer Thompson [mailto:[hidden email]]
*Sent:* Thursday, May 29, 2008 3:23 PM
*To:* Burleson,Joseph A.
*Cc:* [hidden email]
*Subject:* Re: Longitudinal data / data processing
 ----------
From: *hillel vardi* <[hidden email]>
Date: 2008/5/29
To: Jennifer Thompson <[hidden email]>, [hidden email]


Shalom

you can use the fallowing code .

recode symptom1 to symptom10(0 1=0)(2 3 4=1) .
aggregate outfile=* / by  ID / symptom1 to symptom10=sum(symptom1 to
symptom10) .
recode symptom1 to symptom10(1 thru hi=1).

that will give you a file with a line for every Participant and  value 1 for
each symptom if that symptom had value of  2 or higher in any of the
Participant observation.

you can then match that line to the original file if you need it .

Hillel Vardi
BGU



Jennifer Thompson wrote:

>
----------
From: *Carsten Pauck* <[hidden email]>
Date: 2008/5/30
To: [hidden email]
Cc: Jennifer Thompson <[hidden email]>


Hello,
another idea is to restructure the dataset. As a result you have one
row for every patient (=id):
You start with the data:
   ID        DATE                SYMA    SYMB    SYMC
   1,00 01-MAY-1998     ,00    1,00     ,00
   1,00 01-MAY-1999    1,00    2,00     ,00
   1,00 01-MAY-2000    2,00    2,00    3,00
   2,00 02-MAY-1998     ,00     ,00     ,00
   2,00 02-MAY-1999    2,00    3,00    2,00
   3,00 03-MAY-1998    2,00    4,00    2,00
   3,00 03-MAY-1999    2,00    2,00    3,00
   3,00 03-MAY-2000    1,00    2,00    3,00
   3,00 03-MAY-2001    1,00     ,00     ,00
And you restructure the data: Data -> restructure ... which will generate
the
following syntax:
SORT CASES BY id .
CASESTOVARS
/ID = id
/GROUPBY = VARIABLE .
Which will transform the data in switching vars to cases.

syma.1  syma.2 syma.3 syma.4    symb.1  etc
   1,00 01-MAY-1998 01-MAY-1999 01-MAY-2000 .                      ,00
 1,00    2,00 .    1,00    2,00    2,00 .     ,00     ,00    3,00 .
   2,00 02-MAY-1998 02-MAY-1999 . .
,00    2,00 . .     ,00    3,00 . .     ,00    2,00 . .
   3,00 03-MAY-1998 03-MAY-1999 03-MAY-2000 03-MAY-2001    2,00    2,00
 1,00 1,00    4,00    2,00    2,00     ,00    2,00    3,00    3,00     ,00


Now it is possible to analyse the data in the way you are used to (count,
etc ..).
Carsten.


2008/5/29 Jennifer Thompson <[hidden email]>:

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

[Auberth Hurtado]

Hi,

Somebody knows the way to convert all the string variables in my data editor
to 255 width (syntax, script, python)?

I remember i used a script, but don't find it.


Thanks.

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

[Peck, Jon]

If you have SPSS 16, you can use the ALTER TYPE command.
ALTER TYPE ALL (A=A255).

That means any variables with A format (any length) become A255.

HTH,
Jon Peck

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Auberth Hurtado
Sent: Friday, May 30, 2008 11:18 AM
To: [hidden email]
Subject: [SPSSX-L] Convert all string var to 255 character

Hi,

Somebody knows the way to convert all the string variables in my data editor
to 255 width (syntax, script, python)?

I remember i used a script, but don't find it.


Thanks.

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

[Richard Ristow]

At 12:46 PM 5/29/2008, Jennifer Thompson wrote:

>I am working on a longitudinal data set that consists of psychiatric
>symptoms, coded on a 5-point (0-4) scale.  The data set is 'long'
>with a separate case for each observation.  The number of
>observations per participant is variable.  Each participant has a
>unique ID code (combination of letters and digits) and a date is
>recorded for each interview.
>
>I want to create a new variable (e.g. 'ever_panic') that is coded
>'1' if a given symptom (e.g. 'panic') is scored 2 or higher for
>_any_ of the interviews with an individual participant.

ViAnn Beadle is right: AGGREGATE is the prime tool for summarizing in
a 'long' dataset.

I'd use MAX rather than PGT.  Like this (not tested):

AGGREGATE
   /OUTFILE=* MODE=ADDVARIABLES
   /BREAK=Participant_ID
   /ever_panic = MAX(panic).

Now, 'ever_panic' is the worst panic score observed for the patient,
which may do you just fine. Or, you can change to 'yes/no/' with RECODE:

RECODE ever_panic
     (MISSING   = 9)
     (2 THRU HI = 1)
     (ELSE      = 2).

(Note the special clause for missing values; omit it, and you'll get
misleading results when there's any patient without data.)

Now, you write,

>I can only think of tortuously long-winded ways to do this, which is
>not ideal because the process will need to be repeated for 30+ items!

I'm afraid that this won't be as compact as one would like. That is,
you'll need a line in the AGGREGATE for every symptom:


AGGREGATE
   /OUTFILE=* MODE=ADDVARIABLES
   /BREAK=Participant_ID
   /ever_panic   = MAX(panic)
   /ever_depress = MAX(depress)
   /ever_angry   = MAX(angry)
   ...

At least the RECODE stays simple:

RECODE ever_panic TO ever_angry
     (MISSING   = 9)
     (2 THRU HI = 1)
     (ELSE      = 2).

Does this get you nearer where you want to go?

-Best of luck to you,
  Richard

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