So, you're saying that the crosstabs capability is critical. I'm
sure you have a point, I don't use crosstabs much. I'll have to
give that some thought as to whether there is an easy workaround for
that.
-Alan
On 1/8/2015 10:54 AM, ftr wrote:
In fact the Multiple response procedure is particular useful
because stats programs are based on the statistical independence
of observations whereas in survey research you often have multiple
response sets when the same respondent has more than one answer to
a question, i.e. the cases are statistically dependent.
This is why I vote for the implementation of the mult response
proc , for practical reasons and to increase the attractiveness of
PSPP for larger audiences .
Did you already work with multiple response questions ?
Take the example of drinks and age.
Usually you have one answer for the question: what do you drink ?
But in reality you drink Coke as well as water, beer, but not
soda.
So the same person has several answers for the same question.
Now differentiate that by 3 age groups:
The table in MULT RESPONSE can show how many times the same cases
(=persons) in the low, the intermediate, the high age group drink
drink Coke AS WELL AS the another drink beer, either in percentage
of cases (% of persons in the low age group drink Coke, % beer,
etc.) or how many beer drinks are in the this age group.
Better than the Youtube video the text
of John Hall (see p.7) can provide you an idea.
http://surveyresearch.weebly.com/uploads/2/9/9/8/2998485/3.3.2a1__spss_15___first_exercise_in_multiple_response.pdf
- ftr
On 08/01/2015 17:21, Alan Mead wrote:
I've used SPSS to analyze multiple response data for years
(decades, actually) but never used MULT RESPONSE. I was curious
what I was missing, so I watched this video: https://www.youtube.com/watch?v=-toBCDscCwQ
and I'm still a bit confused. You get the same data by running
frequencies on the four variables independently, right?
If each response is optional, then one thing that is a bit of a
PITA is detecting non-response, but that's not a big deal. For
example, if the four possible responses to Q12 are encoded 1/0
in Q12A, Q12B, Q12C, and Q12D, then you can do this:
count Q12MISS = Q12A A12B Q12C Q12D (1).
execute.
Everyone with Q12MISS=0 didn't respond to the question. For some
questions, this is more important than individual responses
(other times not).
I'm not arguing against including it in PSPP, I'm just curious
why it's an issue because it seems like it's really, really easy
to get along without. What am I missing?
BTW, there is another issue of multiple responses that DOESN'T
work this way. When you have a test question labeled "Mark all
that apply" and if your scoring is all or nothing then it's
actually easier to handle this as a string. If they marked A, B
and E on Q12, you encode their response as 'ABE'. Later you
score it: "recode Q12 ('ABC'=1) (else=0) into Q12.Scored." If
you're going to give partial credit for individual responses,
it's usually easier to enter the individual responses as
independent variables, but you could create them using string
functions. So, again, SPSS without MULT RESPONSE seems
perfectly adequate and MULT RESPONSE doesn't actually handle all
multiple-responses situations.
-Alan
On 1/8/2015 8:22 AM, Matthias Faeth
wrote:
I would support that. Multi Response is the one
procedure that lets me stick to SPSS. I'm not a progammer
but would help with testing and comparing.
_______________________________________________
Pspp-users mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/pspp-users
--
Alan D. Mead, Ph.D.
President, Talent Algorithms Inc.
science + technology = better workers
+815.588.3846 (Office)
+267.334.4143 (Mobile)
http://www.alanmead.org
Announcing the Journal of Computerized Adaptive Testing (JCAT), a
peer-reviewed electronic journal designed to advance the science and
practice of computerized adaptive testing: http://www.iacat.org/jcat
_______________________________________________
Pspp-users mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/pspp-users
--
Alan D. Mead, Ph.D.
President, Talent Algorithms Inc.
science + technology = better workers
+815.588.3846 (Office)
+267.334.4143 (Mobile)
http://www.alanmead.org
Announcing the Journal of Computerized Adaptive Testing (JCAT), a
peer-reviewed electronic journal designed to advance the science and
practice of computerized adaptive testing: http://www.iacat.org/jcat
|