Have an idea?

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

Choosing a design (Complete Enumeration, Balanced Overlap...)

Hello,

 my investigation consists of 5 attributes (3 attributes with 3 levels, 2 attributes with 4 levels). There are also some prohibited combinations.

The respondents can choose between two options. (No non-option).

I decided to choose the Complete Enumeration. In SSI Web there`a button to klick on: "Show Levels of First attributes no more than Once per task". What does it mean, should I select this option or not?

Is the complete enumeration the correct design for my investigation?

I am looking forward to a helpful response.

Thanks.
asked May 20, 2014 by anonymous

1 Answer

+1 vote
Choosing between Complete Enumeration, Shortcut, Balanced Overlap, or Random really depends on your goals.

Complete Enumeration and Shortcut will both ensure (unless you have prohibitions) that each level doesn't repeat itself within the same choice task (unless it is forced to make enough concepts to show within the same task).  That is called "minimal overlap"--to not repeat a level within the same task.  

By default, we now in the software pre-select "Balanced Overlap" for you...which leads to a modest degree of level repeating (only some of the attributes will repeat a level within the task).  We do this because of two reasons seen important to most researchers:

1.  When you allow some levels to sometimes repeat within the same choice task, the precision of interaction effects is increased quite a bit...without much loss to precision of main effects.

2.  If each level only appears once per task, if there is a critical "must-have" level for a respondent, this can lead to very superficial responses--the respondent just picks the one level within each task that is the most "must-have".  However, for that respondent if we can sometimes show choice tasks where both concepts have that same most important, must-have level, then we have a chance to observe how that respondent evaluates attributes beyond that first attribute level.

So, in general, when in doubt, pick "Balanced Overlap" for your design method.  But, make sure to perform the Test Design procedure (especially the advanced test design within SSI Web) to see the efficiency scores and expected precision for your attribute levels.  The documentation contains more details about what to look for (press F1 at any dialog for Help).

Because you said you are doing prohibitions between some attributes, please recognize:

1.  Your design efficiency may be compromised (see Test Design results to check that)

2.  You cannot specify an interaction effect when estimating utilities between two attributes involved in a prohibition.  This may be a limitation to you, so consider whether this matters.
answered May 20, 2014 by Bryan Orme Platinum Sawtooth Software, Inc. (132,290 points)
...