Need help excluding a value when computing a variable

[Suzanne Hayes]

I am trying to sum up a series of likert scores to create a new variable but need to excluded any that were answered "Chose not to answer" which were coded 99.

How would I use the transform/compute variable function?  

These are my variables:

Sum (CP_EX_1,CP_EX_2,CP_EX_3,CP_IN_1,CP_IN_2,CP_IN_3,CP_RE_1,CP_RE_2,CP_RE_3,CP_TR_1,CP_TR_2,CP_TR_3)

My if statement seems to be the problem.  I thought that I could "or" to create a set....

if (CP_EX_1 | CP_EX_2 | CP_EX_3 | CP_IN_1 | CP_IN_2 | CP_IN_3 | CP_RE_1 | CP_RE_2 | CP_RE_3 | CP_TR_1 | CP_TR_2 | CP_TR_3) ~= 99


Is there a better way to do this?

S. Hayes
Empire State College

[Maguin, Eugene]

Suzanne,

 

You should have 99 set as a missing value for all variables to which it applies.

 

Gene Maguin

---

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Suzanne Hayes
Sent: Friday, June 24, 2011 3:09 PM
To: [hidden email]
Subject: Need help excluding a value when computing a variable

I am trying to sum up a series of likert scores to create a new variable but need to excluded any that were answered "Chose not to answer" which were coded 99.

How would I use the transform/compute variable function?  

These are my variables:

Sum (CP_EX_1,CP_EX_2,CP_EX_3,CP_IN_1,CP_IN_2,CP_IN_3,CP_RE_1,CP_RE_2,CP_RE_3,CP_TR_1,CP_TR_2,CP_TR_3)

My if statement seems to be the problem.  I thought that I could "or" to create a set....

if (CP_EX_1 | CP_EX_2 | CP_EX_3 | CP_IN_1 | CP_IN_2 | CP_IN_3 | CP_RE_1 | CP_RE_2 | CP_RE_3 | CP_TR_1 | CP_TR_2 | CP_TR_3) ~= 99


Is there a better way to do this?

S. Hayes
Empire State College

[Swank, Paul R]

I would not use a sum if you have missing data as the missingness will be related to the sum. What you do depends on why the data is missing. Imputation is one solution. Using a mean instead of a sum is another although that is more likely to bias results if the missing is not missing completely at random.

 

Dr. Paul R. Swank,

Professor

Children's Learning Institute

University of Texas Health Science Center-Houston

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Suzanne Hayes
Sent: Friday, June 24, 2011 2:09 PM
To: [hidden email]
Subject: Need help excluding a value when computing a variable

 

I am trying to sum up a series of likert scores to create a new variable but need to excluded any that were answered "Chose not to answer" which were coded 99.

How would I use the transform/compute variable function?  

These are my variables:

Sum (CP_EX_1,CP_EX_2,CP_EX_3,CP_IN_1,CP_IN_2,CP_IN_3,CP_RE_1,CP_RE_2,CP_RE_3,CP_TR_1,CP_TR_2,CP_TR_3)

My if statement seems to be the problem.  I thought that I could "or" to create a set....

if (CP_EX_1 | CP_EX_2 | CP_EX_3 | CP_IN_1 | CP_IN_2 | CP_IN_3 | CP_RE_1 | CP_RE_2 | CP_RE_3 | CP_TR_1 | CP_TR_2 | CP_TR_3) ~= 99


Is there a better way to do this?

S. Hayes
Empire State College

[Jon K Peck]

To elaborate further.
missing values <all your variables> (99).
compute sum = sum(x,y,z etc).

will do it.

If you run something like
compute xory = x|y.
you will see that this doesn't make sense if the variables are not true/false (1,0).


Jon Peck
Senior Software Engineer, IBM
[hidden email]
new phone: [phone]




From:        Gene Maguin <[hidden email]>
To:        [hidden email]
Date:        06/24/2011 01:21 PM
Subject:        Re: [SPSSX-L] Need help excluding a value when computing a              variable
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

Suzanne,
 
You should have 99 set as a missing value for all variables to which it applies.
 
Gene Maguin

---

From: SPSSX(r) Discussion [[hidden email]] On Behalf Of Suzanne Hayes
Sent: Friday, June 24, 2011 3:09 PM
To: [hidden email]
Subject: Need help excluding a value when computing a variable


I am trying to sum up a series of likert scores to create a new variable but need to excluded any that were answered "Chose not to answer" which were coded 99.

How would I use the transform/compute variable function?  

These are my variables:

Sum (CP_EX_1,CP_EX_2,CP_EX_3,CP_IN_1,CP_IN_2,CP_IN_3,CP_RE_1,CP_RE_2,CP_RE_3,CP_TR_1,CP_TR_2,CP_TR_3)

My if statement seems to be the problem.  I thought that I could "or" to create a set....

if (CP_EX_1 | CP_EX_2 | CP_EX_3 | CP_IN_1 | CP_IN_2 | CP_IN_3 | CP_RE_1 | CP_RE_2 | CP_RE_3 | CP_TR_1 | CP_TR_2 | CP_TR_3) ~= 99


Is there a better way to do this?

S. Hayes
Empire State College

[Richard Ristow]

At 03:09 PM 6/24/2011, Suzanne Hayes wrote:

>I am trying to sum up a series of likert scores to create a new
>variable but need to excluded any that were answered "Chose not to
>answer" which were coded 99.
>
>How would I use the transform/compute variable function?
>
>These are my variables:
>
>Sum  (CP_EX_1,CP_EX_2,CP_EX_3,CP_IN_1,CP_IN_2,CP_IN_3,
>       CP_RE_1,CP_RE_2,CP_RE_3,CP_TR_1,CP_TR_2,CP_TR_3)

It sounds like you want

MISSING VALUES CP_EX_1 TO CP_TR_3 (99).

before the SUM. If the variables aren't contiguous in the file,
you'll have to give the whole list in the MISSING VALUES statement:

MISSING VALUES CP_EX_1 CP_EX_2 CP_EX_3 CP_IN_1 CP_IN_2 CP_IN_3
                CP_RE_1 CP_RE_2 CP_RE_3 CP_TR_1 CP_TR_2 CP_TR_3
                (99).

=====================
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]

Suzanne,
Really need more info.
Several possible interpretations here:
1. Are you trying to compute this scale ONLY for cases in which there are NO 99's -missing- for any variable?
2. Are you attempting to compute a SUM over items but NOT include those items with 99 in the sum?
----
Your attempt here:
if (CP_EX_1 | CP_EX_2 | CP_EX_3 | CP_IN_1 | CP_IN_2 | CP_IN_3 | CP_RE_1 | CP_RE_2 | CP_RE_3 | CP_TR_1 | CP_TR_2 | CP_TR_3) ~= 99
would suggest option 1.
Basic rewrite:
*N.B.  Cases with ANY missing values will have System Missing on NewVar!.
DO IF NOT(any(99, CP_EX_1 , CP_EX_2 , CP_EX_3 , CP_IN_1 , CP_IN_2 , CP_IN_3,
                             CP_RE_1, CP_RE_2 ,  CP_RE_3 , CP_TR_1 , CP_TR_2 , CP_TR_3) ).
+  COMPUTE newvar=SUM(CP_EX_1 , CP_EX_2 , CP_EX_3 , CP_IN_1 , CP_IN_2 , CP_IN_3,
                               CP_RE_1, CP_RE_2 , CP_RE_3 , CP_TR_1 , CP_TR_2 , CP_TR_3).
END IF.
Of the variables are contiguous in the file then the following is preferred:
DO IF NOT(ANY(99, CP_EX_1 TO CP_TR_3) ).
+  COMPUTE newvar=SUM(CP_EX_1 TO CP_TR_3).
END IF.

OTOH, if you HAVE declared these variables as having missing values (99) then that will fail.
ie MISSING VALUES CP_EX_1 TO CP_TR_3 (99).
However using VALUE() function will peak under the covers so to speak.
DO IF NOT(ANY(VALUE(99), CP_EX_1 TO CP_TR_3) ).
+  COMPUTE newvar=SUM(CP_EX_1 TO CP_TR_3).
END IF.

If Option 2. then simply .
MISSING VALUES CP_EX_1 TO CP_TR_3 (99).
COMPUTE newvar=SUM(CP_EX_1 TO CP_TR_3).
will suffice but as others have commented, the total value is related to the "missingness", but maybe that's what you want.

Suzanne Hayes wrote

I am trying to sum up a series of likert scores to create a new variable but need to excluded any that were answered "Chose not to answer" which were coded 99.

How would I use the transform/compute variable function?  

These are my variables:

Sum (CP_EX_1,CP_EX_2,CP_EX_3,CP_IN_1,CP_IN_2,CP_IN_3,CP_RE_1,CP_RE_2,CP_RE_3,CP_TR_1,CP_TR_2,CP_TR_3)

My if statement seems to be the problem.  I thought that I could "or" to create a set....

if (CP_EX_1 | CP_EX_2 | CP_EX_3 | CP_IN_1 | CP_IN_2 | CP_IN_3 | CP_RE_1 | CP_RE_2 | CP_RE_3 | CP_TR_1 | CP_TR_2 | CP_TR_3) ~= 99


Is there a better way to do this?

S. Hayes
Empire State College

[John F Hall]

There's a set of tutorials on page "3.5  Derived variables (COUNT and COMPUTE)" of my website which covers summing Likert type scales (including reversing codes if necessary) with fully worked examples.

http://surveyresearch.weebly.com/block-3-analysing-two-variables-and-sometimes-three.html

You can also check how many items are missing by:

COUNT x = <varlist> (99).
FREQ x .

. . where <varlist> can be all 12 in your list or, judging by your varnames, separate sets of 3.

If your items are coded eg 1 - 5, and you're adding them together to form a scale, it's sometimes advisable to create a zero point by subtracting the number of items summed.


John F Hall

[hidden email]
www.surveyresearch.weebly.com






-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of David Marso
Sent: 24 June 2011 23:18
To: [hidden email]
Subject: Re: Need help excluding a value when computing a variable

Suzanne,
Really need more info.
Several possible interpretations here:
1. Are you trying to compute this scale ONLY for cases in which there are NO
99's -missing- for any variable?
2. Are you attempting to compute a SUM over items but NOT include those
items with 99 in the sum?
----
Your attempt here:
if (CP_EX_1 | CP_EX_2 | CP_EX_3 | CP_IN_1 | CP_IN_2 | CP_IN_3 | CP_RE_1 |
CP_RE_2 | CP_RE_3 | CP_TR_1 | CP_TR_2 | CP_TR_3) ~= 99
would suggest option 1.
Basic rewrite:
*N.B.  Cases with ANY missing values will have System Missing on NewVar!.
DO IF NOT(any(99, CP_EX_1 , CP_EX_2 , CP_EX_3 , CP_IN_1 , CP_IN_2 , CP_IN_3,
                             CP_RE_1, CP_RE_2 ,  CP_RE_3 , CP_TR_1 , CP_TR_2
, CP_TR_3) ).
+  COMPUTE newvar=SUM(CP_EX_1 , CP_EX_2 , CP_EX_3 , CP_IN_1 , CP_IN_2 ,
CP_IN_3,
                               CP_RE_1, CP_RE_2 , CP_RE_3 , CP_TR_1 ,
CP_TR_2 , CP_TR_3).
END IF.
Of the variables are contiguous in the file then the following is preferred:
DO IF NOT(ANY(99, CP_EX_1 TO CP_TR_3) ).
+  COMPUTE newvar=SUM(CP_EX_1 TO CP_TR_3).
END IF.

OTOH, if you HAVE declared these variables as having missing values (99)
then that will fail.
ie MISSING VALUES CP_EX_1 TO CP_TR_3 (99).
However using VALUE() function will peak under the covers so to speak.
DO IF NOT(ANY(VALUE(99), CP_EX_1 TO CP_TR_3) ).
+  COMPUTE newvar=SUM(CP_EX_1 TO CP_TR_3).
END IF.

If Option 2. then simply .
MISSING VALUES CP_EX_1 TO CP_TR_3 (99).
COMPUTE newvar=SUM(CP_EX_1 TO CP_TR_3).
will suffice but as others have commented, the total value is related to the
"missingness", but maybe that's what you want.



Suzanne Hayes wrote:

>
> I am trying to sum up a series of likert scores to create a new variable
> but need to excluded any that were answered "Chose not to answer" which
> were coded 99.
>
> How would I use the transform/compute variable function?� �
>
> These are my variables:
>
> Sum
> (CP_EX_1,CP_EX_2,CP_EX_3,CP_IN_1,CP_IN_2,CP_IN_3,CP_RE_1,CP_RE_2,CP_RE_3,CP_TR_1,CP_TR_2,CP_TR_3)
>
> My if statement seems to be the problem.�  I thought that I could "or" to
> create a set....
>
> if (CP_EX_1 | CP_EX_2 | CP_EX_3 | CP_IN_1 | CP_IN_2 | CP_IN_3 | CP_RE_1 |
> CP_RE_2 | CP_RE_3 | CP_TR_1 | CP_TR_2 | CP_TR_3) ~= 99
>
>
> Is there a better way to do this?
>
> S. Hayes
> Empire State College
>

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Need-help-excluding-a-value-when-computing-a-variable-tp4522009p4522298.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

[Rich Ulrich]

Years ago, I wanted pro-rated totals so that there could be an
item or two missing without biasing the sum.  That's easy to do :
Use the MEAN.n( ) function instead of SUM, and then multiply by
the total number of items.

Then I noticed that I really preferred to look at the item-average
instead of the sum, because it gave me a natural set of
anchors -- "4.3  on the scale of 1 to 5"  is easier to interpret
than "30 on a scale of 7 to 35 with 7 items"  or "23 on a
scale of 0 to 28." 

--
Rich Ulrich

> Date: Sat, 25 Jun 2011 09:20:29 +0200

> From: [hidden email]
> Subject: Re: Need help excluding a value when computing a variable
> To: [hidden email]
>
> There's a set of tutorials on page "3.5 Derived variables (COUNT and COMPUTE)" of my website which covers summing Likert type scales (including reversing codes if necessary) with fully worked examples.
>
> http://surveyresearch.weebly.com/block-3-analysing-two-variables-and-sometimes-three.html
>
> You can also check how many items are missing by:
>
> COUNT x = <varlist> (99).
> FREQ x .
>
> . . where <varlist> can be all 12 in your list or, judging by your varnames, separate sets of 3.
>
> If your items are coded eg 1 - 5, and you're adding them together to form a scale, it's sometimes advisable to create a zero point by subtracting the number of items summed.
>

 [snip, previous]

[Art Kendall]

I found the same. I found that having the score on the same scale as the response scale was much easier to interpret.

Art

On 6/25/2011 3:03 PM, Rich Ulrich wrote:

Years ago, I wanted pro-rated totals so that there could be an
item or two missing without biasing the sum.  That's easy to do :
Use the MEAN.n( ) function instead of SUM, and then multiply by
the total number of items.

Then I noticed that I really preferred to look at the item-average
instead of the sum, because it gave me a natural set of
anchors -- "4.3  on the scale of 1 to 5"  is easier to interpret
than "30 on a scale of 7 to 35 with 7 items"  or "23 on a
scale of 0 to 28." 

--
Rich Ulrich

> Date: Sat, 25 Jun 2011 09:20:29 +0200
> From: [hidden email]
> Subject: Re: Need help excluding a value when computing a variable
> To: [hidden email]
>
> There's a set of tutorials on page "3.5 Derived variables (COUNT and COMPUTE)" of my website which covers summing Likert type scales (including reversing codes if necessary) with fully worked examples.
>
> http://surveyresearch.weebly.com/block-3-analysing-two-variables-and-sometimes-three.html
>
> You can also check how many items are missing by:
>
> COUNT x = <varlist> (99).
> FREQ x .
>
> . . where <varlist> can be all 12 in your list or, judging by your varnames, separate sets of 3.
>
> If your items are coded eg 1 - 5, and you're adding them together to form a scale, it's sometimes advisable to create a zero point by subtracting the number of items summed.
>
 [snip, previous]

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