6 Replies Latest reply on Aug 18, 2016 8:59 AM by Tom W

    How do I work with a select group of respondents?

    David Roussain

      I need some basic help on how to work with a select group of respondents in a community survey.    I can't figure out how to make graphs based on a subset of respondents to my survey.

       

      One question in my survey is "Do you want to improve the clubhouse"   (Y/N)

       

      I would like to have one slide that focuses only on this group (or is it a set?).    The slide would have responses to other questions FOR THIS GROUP only. 

       

      Real english version --> For respondents who said they want to improve the clubhouse, here are their answers to the following three questions, i.e. Q1 Q7 Q14 etc.

       

      How do I do this?   Should I create a set or group?   How do I make the % responses to the questions as a % of this group only?

       

      Thanks again - just a few pointer to get me started!

       

      D

        • 1. Re: How do I work with a select group of respondents?
          Tom W

          Hi David,

          This will depend entirely on how your data is structured. Do you have one column per question or one row per question?
          If it's a column based approach, just drag the clubhouse question column into the filters shelf, set it accordingly and start to build out a view on their other responses.

          • 2. Re: How do I work with a select group of respondents?
            David Roussain

            Thanks Tom for replying to my question.

             

            I have set up my data table with four fields - survey ID, survey question, and survey response.    

             

            For every survey, I have 17 questions.  There are 76 responses.   The data table therefore has = 1292 (based on 17 * 76)  total rows.

             

            My goal is "for every Y to question 6", what are the answers to three or four other questions?

             

            Suggestions?

             

            Thanks!

            • 3. Re: How do I work with a select group of respondents?
              Tom W

              I work with Survey Data *all the time* and the structure you're using isn't really optimal as it makes what you're trying to do very hard.

              I usually preprocess my data into related tables like 'Respondent' - where I store all field related to the respondent themselves like who they are, where they are etc. One column per attribute / field.

              A quicker approach to this might be to use the pivot options in Tableau, but then you'll end up with a very wide dataset - see https://onlinehelp.tableau.com/current/pro/online/mac/en-us/pivot.html. Additionally, here's an older KB article on working with survey data which speaks about some of the challenges but uses quite an old approach to pivoting the data - Working with Survey Data | Tableau Software . I think it's worth a read, but don't apply the actual learnings within using their methods, use the native pivot method discussed above.

               

              With all that being said, it's possible to get what you want using a Level of Detail calculation. I've created a LOD calc as below which effectively counts/sums the number of times a [Response ID] has a value of Y in Question 6. I can then use this in the IF statement to customize my output and build a filter based on that;

               

               

              IF

              {FIXED [Response ID]: SUM(IF Question=6 and [Response]="Y" then 1 else 0 END)} >0

              THEN 'Include'

              ELSE 'Exclude'

              END

              2 of 2 people found this helpful
              • 4. Re: How do I work with a select group of respondents?
                David Roussain

                Hello Tom ---

                 

                Thank you again for responding.    This may be helpful -- I also have the survey results in an excel file with one row per respondent and each column records the answer to each question.   Hence Column1 = Survey ID, Column 2 = Q1 Column 3 = Q2 etc  (or should I swap the rows and columns?)

                 

                Should I use this data table instead?   The reason I didn't is I watched a Tableau guru on YouTube that recommended converting my (simple) table into a horrendously long file using the Tableau file shaper.   So I did it, thinking it was better.   But now everything seems a challenge.

                 

                So long story short, I can use the shorter excel file (and happy to do so) .

                 

                Recommendations on how to work with this file?

                • 5. Re: How do I work with a select group of respondents?
                  Matthias Goossens

                  Hey David,

                   

                  I would always recommend to work with the following table structure:

                   

                  Column 1 = Subject

                  Column 2 = Question (Q1, Q2, Q3,..)

                  Column 3 = Answers (1,5,3,2)

                   

                   

                  But if you are interested about what Tableau thinks about survey data, then you can always read the following pdf

                  Visualizing Survey Data | Tableau Software

                   

                  Regards,

                  Matthias

                  • 6. Re: How do I work with a select group of respondents?
                    Tom W

                    David,

                    Both approaches have limitations. I.e. multiple select questions or question with multiple options don't lend themselves very well to the 'wide' table approach, but they do work well with the 'deep' table approach.

                     

                    If you use the wide approach, it should be straight forward enough for you to drop the 'Question 6' dimension onto the filter and go about your way.

                    If you use the deep approach, the calculation I've attached above should be adequate.

                     

                    My recommendation, work through both approaches and see where the limitations lie. There isn't really an answer to which is 'better' because it depends on what you're capable of implementing and using.