Response patterns in matrix questions

[emma78]

Hi,
   I have a data set with three sets of 9 items (x1 to x9),  (x10 to x19) and  (x20 to x29) each with 5 response options (1 through 5) for data from 500 participants. Each set stands for one matrix question.  
I would like to identify respondents which response patterns over those 3 matrix questions.


Do you have any hint on this?

Best regards
Emma

[Maguin, Eugene]

Please, oh please, don't make us guess what this ('... identify respondents which response patterns over those 3 matrix  
questions') means. Give us some examples. Explain 'matrix question' and how one matrix question generates 9 items.
Gene Maguin




-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of emma78
Sent: Wednesday, July 29, 2015 12:36 PM
To: [hidden email]
Subject: Response patterns in matrix questions

Hi,
   I have a data set with three sets of 9 items (x1 to x9),  (x10 to x19) and  (x20 to x29) each with 5 response options (1 through 5) for data from 500 participants. Each set stands for one matrix question.  
I would like to identify respondents which response patterns over those 3 matrix questions.


Do you have any hint on this?

Best regards
Emma





--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Response-patterns-in-matrix-questions-tp5730319.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

[David Marso]

Maguin, Eugene wrote

Please, oh please, don't make us guess what this ('... identify respondents which response patterns over those 3 matrix  
questions') means. Give us some examples. Explain 'matrix question' and how one matrix question generates 9 items.
Gene Maguin




-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of emma78
Sent: Wednesday, July 29, 2015 12:36 PM
To: [hidden email]
Subject: Response patterns in matrix questions

Hi,
   I have a data set with three sets of 9 items (x1 to x9),  (x10 to x19) and  (x20 to x29) each with 5 response options (1 through 5) for data from 500 participants. Each set stands for one matrix question.  
I would like to identify respondents which response patterns over those 3 matrix questions.


Do you have any hint on this?

Best regards
Emma
Ever watch the movie Cool Hand Luke?






--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Response-patterns-in-matrix-questions-tp5730319.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

[emma78]

Sure 

x1 to x9: these variables stand for 9 statements to which the respondents can give their approval on a scale from 1 to 5
x10-x19: another nine statements
x20-x29: the third block of statements

e.g.
ID 1
x1-x9:      1  1  2  4  5  5  5  5  2  
x10-x19:   1  1  2  4  5  5  5  5  2  
x20-x29:   1  1  1  1  1  1  1  1  1

for the last block there is a pattern and I can check this by computing  a variance of the variables x20-x29.

But also between the first two blocks there is a kind of regularity. The same codes are clicked in the same sequence.

Is there an easy way to see those Ids with those patterns?

I hope this is clearer now!
Thank You!





   

[David Marso]

First off 3 x 9 ~= 29!!!
The following occurs to me but you'll have to work out the manipulations on the MATRIX file.
DATA LIST LIST /ID  x1 TO x27.
BEGIN DATA
1     1  1  2  4  5  5  5  5  2       1  1  2  4  5  5  5  5  2      1  1  1  1  1  1  1  1  1
END DATA.
VARSTOCASES MAKE s1 FROM x1 TO x9 / MAKE s2 FROM x10 TO x18 /MAKE s3 FROM x19 TO  x27 .
SPLIT FILE BY ID.
CORRELATIONS s1 s2 s3 /MATRIX OUT(*).

emma78 wrote

Sure 

x1 to x9: these variables stand for 9 statements to which the respondents can give their approval on a scale from 1 to 5
x10-x19: another nine statements
x20-x29: the third block of statements

e.g.
ID 1
x1-x9:      1  1  2  4  5  5  5  5  2  
x10-x19:   1  1  2  4  5  5  5  5  2  
x20-x29:   1  1  1  1  1  1  1  1  1

for the last block there is a pattern and I can check this by computing  a variance of the variables x20-x29.

But also between the first two blocks there is a kind of regularity. The same codes are clicked in the same sequence.

Is there an easy way to see those Ids with those patterns?

I hope this is clearer now!
Thank You!

[Andy W]

Here is an example using DO REPEAT. The correlations are a clever idea, but this just counts the number of same values. A value of 9 would mean they are all the same between the pair of responses.


**************************************.
DATA LIST LIST /ID  x1 TO x27.
BEGIN DATA
1     1  1  2  4  5  5  5  5  2       1  1  2  4  5  5  5  5  2      1  1  1  1  1  1  1  1  1
END DATA.

COMPUTE Same12 = 0.
COMPUTE Same13 = 0.
COMPUTE Same23 = 0.
DO REPEAT s1 = x1 TO x9 /s2 = x10 TO x18 /s3 = x19 TO x27.
  IF s1 = s2 Same12 = Same12 + 1.
  IF s1 = s3 Same13 = Same13 + 1.
  IF s2 = s3 Same23 = Same23 + 1.
END REPEAT.
**************************************.

[David Marso]

Same idea but a little more compact:

DO REPEAT s1 = x1 TO x9 /s2 = x10 TO x18 /s3 = x19 TO x27.
COMPUTE Same12=SUM( Same12,s1 EQ s2).
COMPUTE Same13=SUM( Same13,s1 EQ s3).
COMPUTE Same23=SUM( Same23,s2 EQ s3).
END REPEAT.

Note the implicit initialization and the use of the Boolean to increment ;-)
SUM is nice in that way as SUM(sysmis,0)=0, SUM(SYSMIS,1)=1, SUM(value,x)=value +x.

Andy W wrote

Here is an example using DO REPEAT. The correlations are a clever idea, but this just counts the number of same values. A value of 9 would mean they are all the same between the pair of responses.


**************************************.
DATA LIST LIST /ID  x1 TO x27.
BEGIN DATA
1     1  1  2  4  5  5  5  5  2       1  1  2  4  5  5  5  5  2      1  1  1  1  1  1  1  1  1
END DATA.

COMPUTE Same12 = 0.
COMPUTE Same13 = 0.
COMPUTE Same23 = 0.
DO REPEAT s1 = x1 TO x9 /s2 = x10 TO x18 /s3 = x19 TO x27.
  IF s1 = s2 Same12 = Same12 + 1.
  IF s1 = s3 Same13 = Same13 + 1.
  IF s2 = s3 Same23 = Same23 + 1.
END REPEAT.
**************************************.

[Art Kendall]

Very often likert type items are written to be items in a likert type scale.

What to do do depends on the substantive meaning of your variables and the substantive meaning of the sets.

Is there a one-to-one relation say, e.g.,  between the third item in the first set, the third item in the second set, and the third item in the third set?

Are the items in s set intended to be repeated measures of a construct, i.e., intended to be items in a summative scale?

What questions are you using the data to answer?

[emma78]

Thanks so far:-)

Your syntax work, but the problem is that I don`t  have only 27 items but over 200. So it would be rather complicated to compute the 'same' variables for all of them. Is there a way to speed this up?


In the current datase we have 200 items for 40 different proposals (5 items  for each proposal).
For each of the proposal we have questions like how comprehensible, innovative etc. it is. We have a scale from 1 up to 10.

people who answers in the same pattern above 40 proposals have to be deleted from the dataset.

Thank you!

[David Marso]

This is the sort of thing people pay me to program for them!
--

emma78 wrote

Thanks so far:-)

Your syntax work, but the problem is that I don`t  have only 27 items but over 200. So it would be rather complicated to compute the 'same' variables for all of them. Is there a way to speed this up?


In the current datase we have 200 items for 40 different proposals (5 items  for each proposal).
For each of the proposal we have questions like how comprehensible, innovative etc. it is. We have a scale from 1 up to 10.

people who answers in the same pattern above 40 proposals have to be deleted from the dataset.

Thank you!

[Rich Ulrich]

VarsToCases takes you from wide format to long, with 40 records per ID.

AGG on ID and the 5 responses, saving the N-per-pattern if you want it.

AGG this result on ID, saving the N-of-patterns -- the number of patterns
for each person.  If you add this var to the end of the previous AGG, you
have all the pattern information you seem to be interested in selecting on.
One pattern umpteen times?  Only a couple of patterns?

- Pardon me if I happened to miss some vital point; I was ignoring this thread
while the problem posed seemed like extensive manipulation that no one
should want to do.

--
Rich Ulrich

> Date: Fri, 31 Jul 2015 04:40:51 -0700

> From: [hidden email]
> Subject: Re: Response patterns in matrix questions
> To: [hidden email]
>
> Thanks so far:-)
>
> Your syntax work, but the problem is that I don`t have only 27 items but
> over 200. So it would be rather complicated to compute the 'same' variables
> for all of them. Is there a way to speed this up?
>
>
> In the current datase we have 200 items for 40 different proposals (5 items
> for each proposal).
> For each of the proposal we have questions like how comprehensible,
> innovative etc. it is. We have a scale from 1 up to 10.
>
> people who answers in the same pattern above 40 proposals have to be deleted
> from the dataset.
>

...

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