Aggregate adjacent endorsement frequency for checking scale item distributions

[Art Kendall]

The syntax attached and below and demonstrates
"Aggregate adjacent endorsement frequency for checking scale item distributions"

*find "Aggregate adjacent endorsement frequency"
   * the goal to flag items that may have peculiar disributions
   * it may be of use when the candidate item pool.
* is larger than the intended length of the final scale.
* The approach is based on the idea that if two adjacent values do not contain 10% of the cases.
* it should be re-examined.
*.
* this demo has 4 steps.
*(1) simulate cases and likert items with no missing values and some forced distribution changes. This is dataset TestData.

*(2) create a table that shows what what will be in a working dataset.
* use OMS to extract the working data set called ValueCounts.
*(3) flag variables for futher examination etc.
*=================================.
*(1) simulate cases and likert items with no missing values.
new file.
input program.
   NUMERIC ID  (F4).
   NUMERIC      item01 TO item10 (F3).
   vector  item=item01 TO item10.
   loop id = 1001 to 1025.
      loop #p = 1 to 10.
         compute item(#p) =RND(RV.UNIFORM(.5,5.5)).
      end loop.
      end case.
   end loop.
   end file.
end input program.
value labels item01 to item10
   1 'Strongly disagree'
   2 'Disagree'
   3 'Neither'
   4 'Agree'
   5 'Strongly agree'.
* force some weird distributions.
recode item02 (3=1)(else=copy).
recode item03 (1=5)(2=4)(else=copy).
recode item04 (1,2=1)(3=5)(else=copy).
recode item05 (1,2,3=1)(4,5=5)(else=copy).
recode item06 (2,3,4=3)(else=copy).
DATASET NAME   TestData.

*(2)  create a table that shows what will be in the working dataset .
********************************** .
* OMS.
DATASET DECLARE  valuecounts.
OMS
  /SELECT TABLES
  /IF COMMANDS=['CTables'] SUBTYPES=['Custom Table']
  /DESTINATION FORMAT=SAV NUMBERED=TableNumber_
   OUTFILE='ValueCounts'.

CTABLES
   /VLABELS    VARIABLES=item01 item02 item03 item04
      item05 item06 item07 item08
      item09 item10
      DISPLAY=BOTH
   /TABLE item01 [COUNT F40.0] + item02 [COUNT F40.0]
      + item03 [COUNT F40.0] + item04 [COUNT F40.0]
      + item05 [COUNT F40.0] + item06 [COUNT F40.0]
      + item07 [COUNT F40.0] + item08 [COUNT F40.0]
      + item09 [COUNT F40.0] + item10 [COUNT F40.0]
   /CLABELS ROWLABELS=OPPOSITE
   /CATEGORIES VARIABLES=item01 item02 item03 item04
      item05 item06 item07 item08
      item09 item10
      ORDER=A KEY=VALUE EMPTY=INCLUDE.
OMSEND.
dataset activate ValueCounts.
RENAME VARIABLES (Count to Count_D =
  StronglyDisagree Disagree Neither Agree StronglyAgree).
*=====================================.
Compute ResponseCount= Sum(StronglyDisagree to StronglyAgree).

numeric Compare1vs2 Compare2vs3 Compare3vs4 Compare4vs5 (Pct4).
numeric Flag1vs2    Flag2vs3    Flag3vs4    Flag4vs5(f1).
do repeat
   compare =  Compare1vs2 to Compare4vs5
   /flag    =  Flag1vs2    to Flag4vs5
   /myvar1  =  StronglyDisagree   to Agree
   /myvar2  =  Disagree   to StronglyAgree.
   compute compare = 100*((myvar1+myvar2) /ResponseCount).
   recode compare(sysmis=0).
   compute flag =compare lt 10.
end repeat.
list Var1 to ResponseCount.
list Compare1vs2 to Flag4vs5.

[David Marso]

Art,
I guess that works fine for people with CTABLES and always 5 categories in their variables.
If you intend this for general use you need to make it into a macro.
Your basic user will likely not know how to correctly modify it.
Expect lots of emails from noobs if/when you position this as a general solution.
Or for that matter if people with older OMSless systems try to use it.

[Richard Ristow]

At 07:12 PM 3/11/2014, David Marso wrote:

>I guess that works fine for people with CTABLES and always 5 categories in
>their variables.

Well, then, I guess I will put forward the solution I posted at
(14:42:45 -0400) today. It does use V2C, AGG, C2V logic, so it
doesn't require the numbers of items or of responses to be specified.
It works better if the names of the items alphabetize in the desired
order: item01--item10 do, item1--item10 do not.

=====================
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 posted a similar solution which is pretty jugular.

A bit less follow up work than Richard's.

On Tue, Mar 11, 2014 at 7:07 PM, Richard Ristow <[hidden email]> wrote:

At 07:12 PM 3/11/2014, David Marso wrote:

I guess that works fine for people with CTABLES and always 5 categories in
their variables.

Well, then, I guess I will put forward the solution I posted at (14:42:45 -0400) today. It does use V2C, AGG, C2V logic, so it doesn't require the numbers of items or of responses to be specified. It works better if the names of the items alphabetize in the desired order: item01--item10 do, item1--item10 do not.

[Art Kendall]

The set of syntax the list helped me work out was a demo designed more to communicate the concept rather than provide an efficient way to use it.


On 3/11/2014 8:21 PM, David Marso [via SPSSX Discussion] wrote:

I posted a similar solution which is pretty jugular.

A bit less follow up work than Richard's.

On Tue, Mar 11, 2014 at 7:07 PM, Richard Ristow <[hidden email]> wrote:

At 07:12 PM 3/11/2014, David Marso wrote:

I guess that works fine for people with CTABLES and always 5 categories in
their variables.

Well, then, I guess I will put forward the solution I posted at (14:42:45 -0400) today. It does use V2C, AGG, C2V logic, so it doesn't require the numbers of items or of responses to be specified. It works better if the names of the items alphabetize in the desired order: item01--item10 do, item1--item10 do not.

---

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

http://spssx-discussion.1045642.n5.nabble.com/Aggregate-adjacent-endorsement-frequency-for-checking-scale-item-distributions-tp5724838p5724841.html

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

[Bruce Weaver]

For the benefit of anyone reading the archives in future, here are the more general solutions to Art's problem referred to below.  

Richard's solution:  http://spssx-discussion.1045642.n5.nabble.com/How-to-assemble-a-datset-with-with-Likert-or-similar-items-as-rows-and-values-as-columns-tp5724824p5724828.html

David's solution:  http://spssx-discussion.1045642.n5.nabble.com/How-to-assemble-a-datset-with-with-Likert-or-similar-items-as-rows-and-values-as-columns-tp5724824p5724832.html

Note too that I had to modify the SAVE command in David's MATRIX program to get it to run properly on my system (SPSS 21.0.0.2, 64-bit version; Windoze 7 Professional, SP1):

http://spssx-discussion.1045642.n5.nabble.com/How-to-assemble-a-datset-with-with-Likert-or-similar-items-as-rows-and-values-as-columns-tp5724824p5724833.html

HTH.

David Marso wrote

I posted a similar solution which is pretty jugular.
A bit less follow up work than Richard's.


On Tue, Mar 11, 2014 at 7:07 PM, Richard Ristow <[hidden email]>wrote:

> At 07:12 PM 3/11/2014, David Marso wrote:
>
>  I guess that works fine for people with CTABLES and always 5 categories in
>> their variables.
>>
>
> Well, then, I guess I will put forward the solution I posted at (14:42:45
> -0400) today. It does use V2C, AGG, C2V logic, so it doesn't require the
> numbers of items or of responses to be specified. It works better if the
> names of the items alphabetize in the desired order: item01--item10 do,
> item1--item10 do not.
>
>

[Art Kendall]

I do not intend the demo to be a general solution. 
I intend the demo to communicate what is meant by "Aggregate adjacent endorsement frequency".

If I were to be demonstrated in an animated clip. the CTABLE could be generated via the menus and pasted in the syntax. Then the OMS commands could be inserted around that syntax.
However, I am not convinced that this is a very useful tool for screening potential scale items.
On 3/11/2014 7:12 PM, David Marso [via SPSSX Discussion] wrote:

Art,
I guess that works fine for people with CTABLES and always 5 categories in their variables.
If you intend this for general use you need to make it into a macro.
Your basic user will likely not know how to correctly modify it.
Expect lots of emails from noobs if/when you position this as a general solution.

---

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

http://spssx-discussion.1045642.n5.nabble.com/Aggregate-adjacent-endorsement-frequency-for-checking-scale-item-distributions-tp5724838p5724839.html

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