Conjoint Analysis Case Study
The project Discount.Q, which is by default installed on your computer at C:\Program Files\Q\Examples, contains an Experiment question called Ranking-Based Conjoint, from an experiment in which respondents ranked 16 hypothetical discount cards. Each of the hypothetical discount cards varied in terms of the following attributes:
- Coverage, which refers to the proportion of shops providing a discount, with levels: 10%, 20%, 30% and 40%.
- Price, which was the annual fee for the cards, with levels: $30, $45, $60 or $75.
- Linked, which was an option for a logo to be provided on an existing credit card, rather than an additional card being required.
- Warranty, which provided an extended warranty on goods purchased.
- Protection, which provided 90 days theft protection insurance.
Variables 212 to 227 contain 16 variables, one for each of the alternatives (i.e., the 16 hypothetical products). The values of the variables indicate the ranking: the higher the value, the higher the appeal. All 16 variables have different labels; this tells Q that the data contains ranks and that there are 16 different alternatives. The next 16 variables contain the Coverage attribute levels corresponding to the 16 rankings; each variable contains the Label of Coverage, which tells Q that the variables all relate to the same attribute. The next 16 variables relate to the Price Variable, etc.
In the project Discount.Q, the Experiment Choice-Based Conjoint - 1 choice set is from a single choice set containing 16 alternatives (i.e., the assumption is that the respondent has chosen 1 hypothetical product from amongst 16). That is, only a single variable is required to indicate respondents’ preferences. There are 16 variables for each attribute, where each represents the attributes of one of the 16 alternatives. See here for more information about the setup of the data file.
The question Choice-Based Conjoint - 4 choice sets is from 4 choice sets, each containing 4 alternatives. The variables representing the attributes are ordered such that they first contain the variables for each alternative in the first choice set, then the variables for each alternative in the second choice set, and so on. That is, variables 489 to 492 contain the coverage attributes for the first choice set, 493 to 496 for the second choice set, and so on.
The two examples, Choice-Based Conjoint - 1 choice set and Choice-Based Conjoint - 4 choice sets are actually from a ranking-based conjoint where new variables have been created to trick Q into analyzing the data as choice-based conjoint. Discrete Choice Experiment Case Study contains an example of choice-based conjoint where the data was collected as choices.
The Availability design question in Discount.Q shows an example of how to set up the data for an availability experiment (i.e., a labeled choice experiment where the labels vary between tasks). The setup is identical to a Repeated Ranking-Based Conjoint, except that additional alternatives containing missing data have been included to illustrate the setup (a fake 5th alternative has been added to the first three choice sets, and a 4th to the final choice set). Note: the fake alternatives are added because the actual data is not from an availability experiment.