Have an idea?

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

Sample size for max diff w.r.t number of concepts

I have 40 concepts in my study.  How  much sample size is recommended to get good base per concept.

We are planning 3x repetition per concept with 5 concepts per screen which comes to 24 screens. How many exposure per concept in total is recommended .

To reduce LOI, if we may use express max diff and may administer with 20 concepts per respondent, how my sample size calculation will effect in this case.
asked Jan 3, 2019 by Saurabh Aggarwal Gold (33,275 points)

1 Answer

0 votes
Saurabh, there is a way you could compute sample size, if you know the size of the maximum standard errors you were willing to accept.  Then

1.  You create a design, create random data (using Excel or the data generator) for 100 respondents and estimate a logit model.  Calculate the the average standard error.  

2. For any other sample size N you can multiply those standard errors by SQRT(100/N) and see what size standard errors to expect.  

3.  Choose the sample size that gives you the size of standard errors you want.  For CBC we recommend standard errors of 0.05 or less, and I don't see a reason that makes it different for MaxDiff.  

You'll likely need more observations if you do Express.  Also, Sparse gives you better data than Express, so for most purposes I'd use Sparse rather than Express.
answered Jan 8, 2019 by Keith Chrzan Platinum Sawtooth Software, Inc. (90,575 points)