A basic workflow for creating an experimental design is to:
- Create a basic standard design.
- Identify ways of improving the design.
A standard basic design
For many problems, a good-enough design is created by:
- Using the balanced overlap algorithm (see Choice of Experimental Design Algorithm). This algorithm is provided by Sawtooth Software, Q, and Displayr.
- Set the number of versions to 10 or more. Ten versions are usually sufficient. If it is straightforward to have more (and this depends on the data collection software), more is better. Designs with less than 10 versions are also OK; most designs used in academic circumstances only have 1 or a small number of versions.
How to improve a design
The design is then evaluated and improved. See the articles in Experimental Design for Choice-Based Conjoint for key considerations that need to be traded off when creating a good design.
Please sign in to leave a comment.