Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • D-efficiency of Full Factorial Design with dcreate

    Arne Risa Hole

    Would you please advise how to calculate the D-efficiency of a full factorial design created by the genfact command using your dcreate module (available from SSC), if that is possible?

    I understand that D-efficiency is a relative statistic whose meaning is dependent upon various characteristics of a specific design. However, I would also like to determine the relative strength of my fractional factorial design by calculating the proportion that the D-efficiency of my fractional factorial design represents of the D-efficiency of my full factorial design.

    Thanks for making dcreate available and for any advice and insight you may offer.

    Red Owl
    Stata/IC version 15.1, Windows 10 (64-bit)

  • #2
    Red Owl

    In general you can use the evaldes command which is part of the dcreate module to calculate the D-efficiency of a DCE design. However, as you say genfact generates the full factorial design which is not a DCE design in itself, it simply provides all of the possible alternatives in a DCE design. What dcreate does is to create choice sets using those alternatives in a way that attempts to maximise the D-efficiency of the DCE design, given a pre-specified model.

    I hope this clarifies.

    Arne

    Comment


    • #3
      Arne Risa Hole

      Thanks so much for your response. Many studies using DCEs (especially marketing studies) report the strength of the design as a percentage based on the proportion of the D-efficiency of the full factorial design that is reflected in the D-efficiency of the fractional factorial design. For example, DCE designs produced by Sawtooth Software (http://www.sawtoothsoftware.com) report this proportion as a percentage.

      I've been asked how to calculate that proportion. Is there a way to accomplish that using your dcreate module from SSC?

      Thanks again.

      Red Owl
      Stata/IC 15.1, Windows 10 (64-bit)

      Comment


      • #4
        Red Owl

        I cannot help with this, I'm afraid, as I'm not familiar with the measure reported by Sawtooth Software and how this is calculated.

        Arne

        Comment


        • #5
          Arne Risa Hole

          Thanks very much.

          For now, I will continue to report that a DCE fractional factorial design created with dcreate is optimally D-efficient within the constraints imposed.

          Red Owl

          Comment


          • #6
            I am going through your presentation available here https://www.stata.com/meeting/nordic...way16_hole.pdf. and I found it very helpful. On slide 11 what does the J in " Minimum number: K/(J - 1)" stand for?

            Comment

            Working...
            X