Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

How can I derive the the d-effiency value of a given design

I am planning a study. Now I have a given experimental design with 3 attributes (4 levels, 4 levels, 8 levels). I have a given number of choice sets. The design is balanced and orthogonal. How can I derive the d-effiency of my design?
asked Sep 6, 2013 by anonymous
edited Sep 6, 2013 by Walter Williams

1 Answer

0 votes
You can set up an identical study in our CBC software, with the identical attribute structure, numbers of tasks, numbers of concepts per task, and number of versions (blocks).

Have CBC software generate a design.  From the Design tab in CBC software, export that design to .CSV file.  Then, open that file and paste your actual design over the CBC-generated design.  Save the file.

Back on the Design tab, import the modified design file.  (Now, CBC software takes your design into its project files as its own.)  Then, from the Design tab, click the Test Design... button and select the Advanced Test.

The Advanced Test allows you to specify the number of respondents you expect to receive, plus the expected % choice of the None alternative.

At the bottom of the Advanced Test report is reported the Relative D-Efficiency.  This is a relative number that depends on the number of tasks and respondents included (among other things).  So, the typical thing to do is to compare the relative D-efficiency for one design to another one with otherwise matching characteristics.

For example, you might first run D-efficiency on the CBC software generated design first.  Then, after importing your new design that was given to you, you could compare the relative D-efficiency of the two designs.  

For example, if the CBC-generated design had a relative D-efficiency of 2745 and the new design you received had a relative D-efficiency of 2970, the the new design is [1-(2970/2745)] = 8% more efficient than the design that CBC software generated.
answered Sep 6, 2013 by Bryan Orme Platinum Sawtooth Software, Inc. (131,990 points)
edited Sep 6, 2013 by Bryan Orme