DO IF Error

[Ujjawal]

Hi Team,

I would like to set cutoff on the basis of number of records in a file. Run syntax only if criteria is met.
Let me build you the background : To generate correlation analysis, data for at least 30 respondents must be included in the file. I tried to set the criteria using DO IF statement. But,i am getting the following error message :
>Error # 4070.  Command name: END IF
>The command does not follow an unclosed DO IF command.  Maybe the DO IF
>command was not recognized because of an error.  Use the level-of-control
>shown to the left of the SPSS Statistics commands to determine the range of
>LOOPs and DO IFs

Following the syntax i am using :

DEFINE CUTOFF ( ) 30 !ENDDEFINE.

GET FILE = 'DataFile'.

Define !Group (cuts=!CHAREND ('/') / options = !CMDEND)

!DO !vnames !IN (!options)

!LET !corr = !quote(!concat(!unquote('Directory\'),'corr','-',!vnames,'.sav'))

GET File = 'Directory\Datafile1.sav'.

Select If !demogs = !vnames.
Execute.

Dataset Name Currfile.

AGGREGATE
  /OUTFILE=*
  /BREAK=
  /SIZE=N.


DO IF SIZE >= Cutoff.

Dataset Activate Currfile.

CORRELATIONS
/VARIABLES= Engagement Drivers
/MISSING=PAIRWISE
/MATRIX = OUT (!corr).

**** HERE I AM GETTING THE ERROR :
>Error # 4070.  Command name: END IF
>The command does not follow an unclosed DO IF command.  Maybe the DO IF
>command was not recognized because of an error.  Use the level-of-control
>shown to the left of the SPSS Statistics commands to determine the range of
>LOOPs and DO IFs. ******

!DOEND

END IF.
EXECUTE.

!ENDDEFINE.
SET MPRINT ON PRINTBACK ON.

!Group cuts=Age /options=1 2.  

It would be very helpful if you suggest any workaround.

Thanks in advance !
Ujjawal Bhandari

[David Marso]

DO IF only allows transformations *not* procedures.
There are work arounds but I'm not going there today!

Ujjawal wrote

Hi Team,

I would like to set cutoff on the basis of number of records in a file. Run syntax only if criteria is met.
Let me build you the background : To generate correlation analysis, data for at least 30 respondents must be included in the file. I tried to set the criteria using DO IF statement. But,i am getting the following error message :
>Error # 4070.  Command name: END IF
>The command does not follow an unclosed DO IF command.  Maybe the DO IF
>command was not recognized because of an error.  Use the level-of-control
>shown to the left of the SPSS Statistics commands to determine the range of
>LOOPs and DO IFs

Following the syntax i am using :

DEFINE CUTOFF ( ) 30 !ENDDEFINE.

GET FILE = 'DataFile'.

Define !Group (cuts=!CHAREND ('/') / options = !CMDEND)

!DO !vnames !IN (!options)

!LET !corr = !quote(!concat(!unquote('Directory\'),'corr','-',!vnames,'.sav'))

GET File = 'Directory\Datafile1.sav'.

Select If !demogs = !vnames.
Execute.

Dataset Name Currfile.

AGGREGATE
  /OUTFILE=*
  /BREAK=
  /SIZE=N.


DO IF SIZE >= Cutoff.

Dataset Activate Currfile.

CORRELATIONS
/VARIABLES= Engagement Drivers
/MISSING=PAIRWISE
/MATRIX = OUT (!corr).

**** HERE I AM GETTING THE ERROR :
>Error # 4070.  Command name: END IF
>The command does not follow an unclosed DO IF command.  Maybe the DO IF
>command was not recognized because of an error.  Use the level-of-control
>shown to the left of the SPSS Statistics commands to determine the range of
>LOOPs and DO IFs. ******

!DOEND

END IF.
EXECUTE.

!ENDDEFINE.
SET MPRINT ON PRINTBACK ON.

!Group cuts=Age /options=1 2.  

It would be very helpful if you suggest any workaround.

Thanks in advance !
Ujjawal Bhandari

[Ujjawal]

Thanks a ton David for your response. Suggest me work arounds when you think it's the best time ;)

[David Marso]

Meanwhile review your code.  You have !END IF and !IFEND swapped.  That however won't have any bearing on the other issue.  Maybe reveal your broader plan.  Perhaps there are alternative approaches (hint MATRIX or gasp python).  

Ujjawal wrote

Thanks a ton David for your response. Suggest me work arounds when you think it's the best time ;)

[Ujjawal]

David, sorry for the late reply. Must apologize for this .
Instead of DO IF i used !IF-!THEN. But, this gives me an error while processing this line

!IF SIZE >= Cutoff !THEN

 ">In an !IF command, the parenthesized logical expression is missing.  !IF must
>be followed by a parenthesized logical expression and a !THEN command.
>This command not executed.

I understand variable SIZE must be a macro expression.

I had spent almost a day to make this code running but unable to crack it. Any help would be highly appreciated !

David, if you remember my query related to matrix posted last week. This query is linked to that problem. I wish to run the analysis by sub group level. I succeeded using !DO loop. Now, i want to take this to the next level by setting minimum respondent level cutoff.

[David Marso]

As I hinted previously.  MATRIX!
MATRIX.
Load data here (say raw data is in X).
DO IF (NROW(X) > !cutoff).
whatever matrix code to do your calculations.
END IF.
END MATRIX.
It is a RPITA to resolve data values in MACRO.
You know I do consulting and training for a living.
My tips on SPSSX-L are really just the tip of an iceberg.

Ujjawal wrote

David, sorry for the late reply. Must apologize for this .
Instead of DO IF i used !IF-!THEN. But, this gives me an error while processing this line

!IF SIZE >= Cutoff !THEN

 ">In an !IF command, the parenthesized logical expression is missing.  !IF must
>be followed by a parenthesized logical expression and a !THEN command.
>This command not executed.

I understand variable SIZE must be a macro expression.

I had spent almost a day to make this code running but unable to crack it. Any help would be highly appreciated !

David, if you remember my query related to matrix posted last week. This query is linked to that problem. I wish to run the analysis by sub group level. I succeeded using !DO loop. Now, i want to take this to the next level by setting minimum respondent level cutoff.

[Ujjawal]

Thank you so much David for taking time out of your schedule.

In this case, matrix calculations are performed on correlation matrix. Hence, NROW doesn't serve the purpose. I mean raw data is in correlation matrix format.

I am sorry if i am missing something silly here.

Just wanted to let you know that i used AGGREGATE to count the number of cases. Is it the right approach? I am trying to assign the number of records to LET statement and then setting criteria on the basis of it.

GET File = 'Datafile.sav'.

Select If Var1 = 2.
Execute.

AGGREGATE
  /OUTFILE=*
  /BREAK=
  /SIZE=N.

!LET !a = SIZE

!IF (!a !GE 30) !THEN

***Correlation analysis if a criteria is met otherwise STOP******

Now, the situation is : NO ERROR but syntax is not considering the number of respondent criteria.

[David Marso]

"!LET !a = SIZE "
And you found to your dismay that MACRO is clueless WRT SIZE!
What I was hinting at was to use MATRIX to compute the correlation matrix and do the bookkeeping .
IIRC when you were reading stacked correlation matrices you were getting silliness for the element names.
(of course there is no way to know how many there are or any sane way to address them within MATRIX so it is rather useless for raw split file matrix input -unless I am missing something-).
Still, at this point you need to lay out the big picture you are trying to paint.  

Ujjawal wrote

Thank you so much David for taking time out of your schedule.

In this case, matrix calculations are performed on correlation matrix. Hence, NROW doesn't serve the purpose. I mean raw data is in correlation matrix format.

I am sorry if i am missing something silly here.

Just wanted to let you know that i used AGGREGATE to count the number of cases. Is it the right approach? I am trying to assign the number of records to LET statement and then setting criteria on the basis of it.

GET File = 'Datafile.sav'.

Select If Var1 = 2.
Execute.

AGGREGATE
  /OUTFILE=*
  /BREAK=
  /SIZE=N.

!LET !a = SIZE

!IF (!a !GE 30) !THEN

***Correlation analysis if a criteria is met otherwise STOP******

Now, the situation is : NO ERROR but syntax is not considering the number of respondent criteria.

[Ujjawal]

Thanks David...Tough time has come . I need to have patience ;)

[David Marso]

Patience is a good thing.  A plan is better.
Take a few steps back and lay out exactly what needs to be done.
Solve each step one at a time.
--

Ujjawal wrote

Thanks David...Tough time has come . I need to have patience ;)

[Fuller, Matthew]

I will be out of the office until March 20th with limited access to e-mail.

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

[Ujjawal]

As you suggested, i took a few steps back. Thought of all the possibilities that i could.

One of these is as follows :
To workaround this problem, i tried to convert values of all the variables to system missing.This serves the purpose. However, it gives errors while executing regression and matrix statements as no values in both independent and dependent variables.

AGGREGATE
  /OUTFILE=* MODE=ADDVARIABLES
  /BREAK=
  /SIZE=N.

DO IF (SIZE LT 30).
RECODE Drivers (ELSE = SYSMIS).
CORRELATIONS
/VARIABLES= Engagement Drivers
/MISSING=PAIRWISE
/MATRIX = OUT ('Directory\corr.sav').


oms
  /select tables
  /if commands =['Regression'] SUBTYPES=['Coefficients']
  /destination format =SAV  outfile ='Directory\Beta.sav'
  /columns sequence =[RALL CALL LALL].

regression
  /dependent Engagement
  /method= enter Drivers.
omsend.

***THEN ALL THE MATRIX STUFF*****

This is the only thing that is spoiling my whole program.

[David Marso]

Back up even further.
I don't want to know what you are doing.  
Because that is obviously problematic.
I want you to lay out what you are trying to do.
--

Ujjawal wrote

As you suggested, i took a few steps back. Thought of all the possibilities that i could.

One of these is as follows :
To workaround this problem, i tried to convert values of all the variables to system missing.This serves the purpose. However, it gives errors while executing regression and matrix statements as no values in both independent and dependent variables.

AGGREGATE
  /OUTFILE=* MODE=ADDVARIABLES
  /BREAK=
  /SIZE=N.

DO IF (SIZE LT 30).
RECODE Drivers (ELSE = SYSMIS).
CORRELATIONS
/VARIABLES= Engagement Drivers
/MISSING=PAIRWISE
/MATRIX = OUT ('Directory\corr.sav').


oms
  /select tables
  /if commands =['Regression'] SUBTYPES=['Coefficients']
  /destination format =SAV  outfile ='Directory\Beta.sav'
  /columns sequence =[RALL CALL LALL].

regression
  /dependent Engagement
  /method= enter Drivers.
omsend.

***THEN ALL THE MATRIX STUFF*****

This is the only thing that is spoiling my whole program.

[David Marso]

Consider the original dilemma: MATRIX was interpreting the SPSS matrix file and parsing out sets of materials which would be very difficult to resolve (even I would abandon all hope...-the code would be too ugly and I don't have enough bleach on hand-).  Consider wiping those assumptions and passing a 'sanitized' version (ie NIX all of the indicators of it being a matrix file from CORR (see AUTORECODE and DELETE VARIABLES)).  Of course you will need to enumerate and process the sub matrices on your own.
That is all I have to say on this.

David Marso wrote

Back up even further.
I don't want to know what you are doing.  
Because that is obviously problematic.
I want you to lay out what you are trying to do.
--

Ujjawal wrote

As you suggested, i took a few steps back. Thought of all the possibilities that i could.

One of these is as follows :
To workaround this problem, i tried to convert values of all the variables to system missing.This serves the purpose. However, it gives errors while executing regression and matrix statements as no values in both independent and dependent variables.

AGGREGATE
  /OUTFILE=* MODE=ADDVARIABLES
  /BREAK=
  /SIZE=N.

DO IF (SIZE LT 30).
RECODE Drivers (ELSE = SYSMIS).
CORRELATIONS
/VARIABLES= Engagement Drivers
/MISSING=PAIRWISE
/MATRIX = OUT ('Directory\corr.sav').


oms
  /select tables
  /if commands =['Regression'] SUBTYPES=['Coefficients']
  /destination format =SAV  outfile ='Directory\Beta.sav'
  /columns sequence =[RALL CALL LALL].

regression
  /dependent Engagement
  /method= enter Drivers.
omsend.

***THEN ALL THE MATRIX STUFF*****

This is the only thing that is spoiling my whole program.

[David Marso]

OK One more:  But thats all folks.
--
GET
  FILE='C:\Program Files\SPSS\Employee data.sav'.
SORT CASES BY JOBCAT.
SPLIT FILE BY JOBCAT.
DATASET DECLARE corrout.
CORRELATIONS
  /MATRIX OUT (corrout)
  /VARIABLES=salary salbegin jobtime prevexp
  /PRINT=TWOTAIL NOSIG
  /MISSING=PAIRWISE.
DATASET ACTIVATE corrout.
SPLIT FILE OFF.
AUTORECODE ROWTYPE_ /INTO RT.
DELETE VARIABLES ROWTYPE_ VARNAME_ .
LIST.
MATRIX.
GET DATA / FILE * / VAR ALL.
PRINT DATA / FORMAT "F8.3".
END MATRIX.

Run MATRIX procedure:

DATA
    1.000 27838.54 14096.05   81.066   85.039    2.000
    1.000 7567.995 2907.474   10.110   95.275    4.000
    1.000  363.000  363.000  363.000  363.000    3.000
    1.000  363.000  363.000  363.000  363.000    3.000
    1.000  363.000  363.000  363.000  363.000    3.000
    1.000  363.000  363.000  363.000  363.000    3.000
    1.000    1.000     .683     .147    -.214    1.000
    1.000     .683    1.000    -.053     .080    1.000
    1.000     .147    -.053    1.000    -.024    1.000
    1.000    -.214     .080    -.024    1.000    1.000
    2.000 30938.89 15077.78   81.556  298.111    2.000
    2.000 2114.616 1341.235    8.487  101.426    4.000
    2.000   27.000   27.000   27.000   27.000    3.000
    2.000   27.000   27.000   27.000   27.000    3.000
    2.000   27.000   27.000   27.000   27.000    3.000
    2.000   27.000   27.000   27.000   27.000    3.000
    2.000    1.000     .077    -.096     .284    1.000
    2.000     .077    1.000    -.543    -.028    1.000
    2.000    -.096    -.543    1.000     .076    1.000
    2.000     .284    -.028     .076    1.000    1.000
    3.000 63977.80 30257.86   81.155   77.619    2.000
    3.000 18244.78 9980.979   10.410   73.260    4.000
    3.000   84.000   84.000   84.000   84.000    3.000
    3.000   84.000   84.000   84.000   84.000    3.000
    3.000   84.000   84.000   84.000   84.000    3.000
    3.000   84.000   84.000   84.000   84.000    3.000
    3.000    1.000     .693     .168     .094    1.000
    3.000     .693    1.000    -.010     .479    1.000
    3.000     .168    -.010    1.000     .091    1.000
    3.000     .094     .479     .091    1.000    1.000

------ END MATRIX -----
Compared to presanitized matrix file:




jobcat ROWTYPE_ VARNAME_     salary   salbegin    jobtime    prevexp

   1   MEAN              27838.5399 14096.0468 81.0661157 85.0385675
   1   STDDEV            7567.99496 2907.47415 10.1104963 95.2746460
   1   N        salary   363.000000 363.000000 363.000000 363.000000
   1   N        salbegin 363.000000 363.000000 363.000000 363.000000
   1   N        jobtime  363.000000 363.000000 363.000000 363.000000
   1   N        prevexp  363.000000 363.000000 363.000000 363.000000
   1   CORR     salary    1.0000000   .6826561   .1466784  -.2140851
   1   CORR     salbegin   .6826561  1.0000000  -.0531836   .0799518
   1   CORR     jobtime    .1466784  -.0531836  1.0000000  -.0244330
   1   CORR     prevexp   -.2140851   .0799518  -.0244330  1.0000000
   2   MEAN              30938.8889 15077.7778 81.5555556 298.111111
   2   STDDEV            2114.61641 1341.23460  8.4867922 101.426492
   2   N        salary   27.0000000 27.0000000 27.0000000 27.0000000
   2   N        salbegin 27.0000000 27.0000000 27.0000000 27.0000000
   2   N        jobtime  27.0000000 27.0000000 27.0000000 27.0000000
   2   N        prevexp  27.0000000 27.0000000 27.0000000 27.0000000
   2   CORR     salary    1.0000000   .0770036  -.0960843   .2835744
   2   CORR     salbegin   .0770036  1.0000000  -.5427114  -.0279289
   2   CORR     jobtime   -.0960843  -.5427114  1.0000000   .0759741
   2   CORR     prevexp    .2835744  -.0279289   .0759741  1.0000000
   3   MEAN              63977.7976 30257.8571 81.1547619 77.6190476
   3   STDDEV            18244.7760 9980.97939 10.4102522 73.2598782
   3   N        salary   84.0000000 84.0000000 84.0000000 84.0000000
   3   N        salbegin 84.0000000 84.0000000 84.0000000 84.0000000
   3   N        jobtime  84.0000000 84.0000000 84.0000000 84.0000000
   3   N        prevexp  84.0000000 84.0000000 84.0000000 84.0000000
   3   CORR     salary    1.0000000   .6926902   .1675453   .0938535
   3   CORR     salbegin   .6926902  1.0000000  -.0100060   .4788105
   3   CORR     jobtime    .1675453  -.0100060  1.0000000   .0910731
   3   CORR     prevexp    .0938535   .4788105   .0910731  1.0000000


Number of cases read:  30    Number of cases listed:  30

David Marso wrote

Consider the original dilemma: MATRIX was interpreting the SPSS matrix file and parsing out sets of materials which would be very difficult to resolve (even I would abandon all hope...-the code would be too ugly and I don't have enough bleach on hand-).  Consider wiping those assumptions and passing a 'sanitized' version (ie NIX all of the indicators of it being a matrix file from CORR (see AUTORECODE and DELETE VARIABLES)).  Of course you will need to enumerate and process the sub matrices on your own.
That is all I have to say on this.

David Marso wrote

Back up even further.
I don't want to know what you are doing.  
Because that is obviously problematic.
I want you to lay out what you are trying to do.
--

Ujjawal wrote

As you suggested, i took a few steps back. Thought of all the possibilities that i could.

One of these is as follows :
To workaround this problem, i tried to convert values of all the variables to system missing.This serves the purpose. However, it gives errors while executing regression and matrix statements as no values in both independent and dependent variables.

AGGREGATE
  /OUTFILE=* MODE=ADDVARIABLES
  /BREAK=
  /SIZE=N.

DO IF (SIZE LT 30).
RECODE Drivers (ELSE = SYSMIS).
CORRELATIONS
/VARIABLES= Engagement Drivers
/MISSING=PAIRWISE
/MATRIX = OUT ('Directory\corr.sav').


oms
  /select tables
  /if commands =['Regression'] SUBTYPES=['Coefficients']
  /destination format =SAV  outfile ='Directory\Beta.sav'
  /columns sequence =[RALL CALL LALL].

regression
  /dependent Engagement
  /method= enter Drivers.
omsend.

***THEN ALL THE MATRIX STUFF*****

This is the only thing that is spoiling my whole program.

[Ujjawal]

Thanks a ton David :) I won't bother you more.

Status : I found a quick fix. When criteria is not met, i let SPSS to run all the analysis and then match the final file with the file wherein variable for respondent size. And then convert all the variables to sysmis using DO IF and Recode.

The problem with the matrix approach is processing the sub matrices when number of records are less than 30. Because you had made them blank SPSS gives errors while processing the statements.

I wish a command like DO IF for procedures :(

[David Marso]

Like I said: Study MATRIX and roll your own when necessary.
MATRIX supports a DO IF -END IF and you can smash any MATRIX code in between so you DO HAVE A CONDITIONAL PROCEDURE.  You just have to roll it yourself .  I thought that was what your original issue was in the first place (you had someone else's MATRIX program and were trying to hammer away with SPLIT FILE and it bit you).
TIP: You might want to venture beyond your little thread here and review what other topics have been posted today ;-)   You perhaps will find a pleasant surprise !!!!!!!

Ujjawal wrote

Thanks a ton David :) I won't bother you more.

Status : I found a quick fix. When criteria is not met, i let SPSS to run all the analysis and then match the final file with the file wherein variable for respondent size. And then convert all the variables to sysmis using DO IF and Recode.

The problem with the matrix approach is processing the sub matrices when number of records are less than 30. Because you had made them blank SPSS gives errors while processing the statements.

I wish a command like DO IF for procedures :(

[ViAnn Beadle]

There is a command like DO IF for procedures. Read up on the FILTER command.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Ujjawal
Sent: Monday, March 18, 2013 1:39 PM
To: [hidden email]
Subject: Re: DO IF Error

Thanks a ton David :) I won't bother you more.

Status : I found a quick fix. When criteria is not met, i let SPSS to run
all the analysis and then match the final file with the file wherein
variable for respondent size. And then convert all the variables to sysmis.

The problem with the matrix approach is processing the sub matrices when
number of records are less than 30. Because you had made them blank SPSS
gives errors while processing the statements.

I wish a command like DO IF for procedures :(



--
View this message in context:
http://spssx-discussion.1045642.n5.nabble.com/DO-IF-Error-tp5718730p5718827.
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

[Art Kendall]

> And then convert all the variables to sysmis.

If you, the user, are doing the transformation then they should be
assigned a USER-MISSING value.

In the long run, it is very important for quality assurance and
debugging to maintain the distinction between
"missing for a known reason"  and
"missing because the system was not able to obey your instructions".

It is usually bad practice to put sysmis on the right side of an
assignment command. The exception would be sometimes in a recode where
sysmis in the input variable is wanted as sysmis on the new variable.

I wonder if there could be a Python app that looked for bad habits like
that.  Perhaps this list could make a list of things that gave a "are
you sure you want to".

Art Kendall
Social Research Consultants

On 3/18/2013 3:38 PM, Ujjawal wrote:

> Thanks a ton David :) I won't bother you more.
>
> Status : I found a quick fix. When criteria is not met, i let SPSS to run
> all the analysis and then match the final file with the file wherein
> variable for respondent size. And then convert all the variables to sysmis.
>
> The problem with the matrix approach is processing the sub matrices when
> number of records are less than 30. Because you had made them blank SPSS
> gives errors while processing the statements.
>
> I wish a command like DO IF for procedures :(
>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/DO-IF-Error-tp5718730p5718827.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

[Ujjawal]

@ Art Kendall: I second that. But the problem is i am unable to find a way to get it run without converting all the variables to symsis. If the number of records in a raw data file is less than 30 this conversion will happen otherwise missing for a known reason.

As David suggested, i have started studying MATRIX-END MATRIX. I hope i will be able to crack it using MATRIX.