A full-profile design presents people with alternatives that show all of the attributes. A partial-profile design only shows a subset of the alternatives. In the example questions below, four of the seven attributes are shown in each question, with the specific attributes being selected chosen to maximize the efficiency of the data collection (i.e., they are efficient designs).
The choice between a full- and partial-profile design comes down to a tradeoff of the following:
- Realism. Full-profile designs are usually more realistic. See Ecological Validity.
- Cognitive difficulty. Partial-profile designs are easier for respondents to complete. With as few as three attributes, this factor is a real consideration. With six or more attributes a partial-profile design should be seriously considered. Most choice models show three or four alternatives (excluding none of these options). A simple rule is that you want the number of alternatives (excluding none of these) multiplied by the number of attributes to be 20 or less. In practice, this rule is often broken, and collecting two alternatives is, from a statistical perspective, considerably worse than collecting three. (This can be assessed by comparing d-error.)
- Sample size. Partial-profile designs require longer questionnaires or more respondents, as each question obtains no information about a subset of the attributes. See Sample Size.