0 Replies Latest reply on Aug 14, 2014 8:38 AM by Laetitia Tlemsani

    Significance test for porportion

    Laetitia Tlemsani

      Hello there,

      I wish to submit a problematic encounter with Tableau on a calculation of significance. I work with a file of survey results as follows:

       

      Year

      Category_1

      Weighted_responses

      Nb_surveys

      2013

      A

      5,057,576

      15,030

      2013

      B

      9,888,532

      36,375

      2013

      Etc.

      2012

      A

      5,477,234

      15,444

      2012

      B

      10,512,112

      36,529

      2012

      Etc.

       

       

      I want to know for each case (A, B, etc.) in each category, if the proportion in 2013 is significantly different from the one in 2012, and I only want to show the difference in point between these two proportions if this difference is indeed significant.

       

      Usually I make it in Excel with a significance test which is calculated as follows:
      Test=(P1 – P2)/sqrt(P(1-P)*(1/N1+1/N2)).

       

      With the figures in the table above, we have for the case A:
      • P1 = 5,057,576 / (5,057,576 + 9,888,532 + ...)
      • P2 = 5,477,234 / (5,477,234 + 10,512,112 + ...)
      • P = (5,057,576 + 5,477,234) / ((5,057,576 + 9,888,532 + ...) + (5,477,234 + 10,512,112 + ...))
      • N1 = 15,030
      • N2 = 15.444

       

      I have to do the exercise on categories that may have more than 50 cases, so I have to use a dynamic formula.

      I don't know how to figure it out in Tableau. Could you help me on this ?

      Thanks a lot !

       

      (by the way I'm french, sorry for my english ! )