SPSS Statistics

 View Only
  • 1.  Multivariate analysis of a LOT of questions

    Posted Mon May 17, 2021 04:09 PM

    Hi all, please excuse a possibly somewhat newbish question. I'm new to stats and SPSS, have worked out enough to do most of the basic things I need but I have a complex scenario that I'm hoping to use SPSS to solve.

    I'm researching 3 different but related job roles. These involve activities that are potentially undertaken by all 3 (but not necessarily so)
    I want to see if there are some activities that are strongly associated with a certain role - both by people in that role and also those in the other 2.

    Question is essentially - do you believe that a person in role A/B/C does activity X

    So I have data from a Qualtrics survey where I have 3 variables for each question. 25 questions (activities) so 75 variables.

    I believe this is a multivariate analysis.

    I would like to be able to say - the top 5 activities for someone in role A are ...
    For activity X, 23% of people in role A think that people in role B do this but 50% of people in role B think that people in role B do this. (etc)

    My data is in wide format - one row for each respondent.

    Any suggestions on a good approach to answer these questions?

    Thanks



    ------------------------------
    Colin Simpson
    ------------------------------


    #SPSSStatistics


  • 2.  RE: Multivariate analysis of a LOT of questions

    IBM Champion
    Posted Mon May 17, 2021 09:31 PM
    It would help to have a sample of the data, but you might start by defining a multiple response set (multiple dichotomies) for each role - three sets.  Then for starters, use CTABLES with the set in the rows and in Categories and Totals, sort by count with descending order.

    Working with these data might be easier if  you restructured the data to long form with three groups.

    --





  • 3.  RE: Multivariate analysis of a LOT of questions

    Posted Wed May 19, 2021 06:30 PM
    Thanks Jon, I'll give that a try. (Uni ethics being what they are, I'm not really keen to share the data itself, even though I know it wouldn't really say much)

    ------------------------------
    Colin Simpson
    ------------------------------