Launch Window for Multiple Tables, Cross-Referenced
Launch the Choice platform by selecting Analyze > Consumer Research > Choice. If your Choice data is in multiple tables, select Multiple Tables, Cross-Referenced from the Data Format menu.
Figure 4.13 Launch Window for Multiple Tables, Cross-Referenced Data Format
Figure 4.13 shows the launch window for Multiple Tables, using Pizza Profiles.jmp as the Profile table.
In the case of Multiple Tables, Cross-referenced, the launch window has three sections:
Profile Data
The profile data table describes the attributes associated with each choice. Each attribute defines a column in the data table. There is a row for each profile. A column in the table contains a unique identifier for each profile. Figure 4.14 shows the Pizza Profiles.jmp data table and a completed Profile Data panel.
Figure 4.14 Profile Data Table and Completed Profile Data Outline
Select Data Table
Select or open the data table that contains the profile data. Select Other to open a file that is not already open.
Profile ID
Identifier for each row of attribute combinations (profile). If the Profile ID column does not uniquely identify each row in the profile data table, you need to add Grouping columns. Add Grouping columns until the combination of Grouping and Profile ID columns uniquely identify the row, or profile.
Grouping
A column which, when used with the Profile ID column, uniquely designates each choice set. For example, if Profile ID = 1 for Survey = A, and a different Profile ID = 1 for Survey = B, then Survey would be used as a Grouping column.
Construct Profile Effects
Add effects constructed from the attributes in the profiles.
For information about the Construct Profile Effects panel, see “Construct Model Effects” in Fitting Linear Models.
Note: The choice model observes the column coding property of continuous profile and subject effects.
Firth Bias-adjusted Estimates
Computes bias-corrected MLEs that produce better estimates and tests than MLEs without bias correction. These estimates also improve separation problems that tend to occur in logistic-type models. See Heinze and Schemper (2002) for a discussion of the separation problem in logistic regression.
Hierarchical Bayes
Uses a Bayesian approach to estimate subject-specific parameters. See Bayesian Parameter Estimates.
Number of Bayesian Iterations
(Applicable only if Hierarchical Bayes is selected.) The total number of iterations of the adaptive Bayes algorithm used to estimate subject effects. This number includes a burn-in period of iterations that are discarded. The number of burn-in iterations is equal to half of the Number of Bayesian Iterations specified on the launch window.
Response Data
The response data table includes a subject identifier column, columns that list the profile identifiers for the profiles in each choice set, and a column containing the preferred profile identifier. There is a row for each subject and choice set. Grouping variables can be used to distinguish choice sets when the data contain more than one group of choice sets. Figure 4.15 shows the Pizza Responses.jmp data table and a completed Response Data panel.
Grouping variables can be used to align choice indices when more than one group is contained within the data.
Figure 4.15 Response Data Table and Completed Responses Data Outline
Select Data Table
Select or open the data table that contains the response data. Select Other to open a file that is not already open.
Profile ID Chosen
The Profile ID from the Profile data table that represents the subject’s selected profile.
Grouping
A column which, when used with the Profile ID Chosen column, uniquely designates each choice set.
Profile ID Choices
The Profile IDs of the set of possible profiles. There must be at least two profiles.
Subject ID
An identifier for the study participant.
Freq
A column containing frequencies. If n is the value of the Freq variable for a given row, then that row is used in computations n times. If it is less than 1 or missing, then JMP does not use it to calculate any analyses.
Weight
A column containing a weight for each observation in the data table. The weight is included in analyses only when its value is greater than zero.
By
Produces a separate report for each level of the By Variable. If more than one By variable is assigned, a separate analysis is produced for each possible combination of the levels of the By variables.
Respondent is allowed to select “None” or “No Choice”
Enters a No Choice Indicator into the model for response rows containing missing values. For the Multiple Tables, Cross-Referenced data format, the No Choice rows must contain (categorical) missing values in the Profile ID Chosen column in the Response Data table. The option appears at the bottom of the Response Data panel.
Subject Data
The subject data table is optional and depends on whether you want to model subject effects. The table contains a column with the subject identifier used in the response table, and columns for attributes or characteristics of the subjects. You can put subject data in the response data table, but you should specify the subject effects in the Subject Data outline. Figure 4.16 shows the Pizza Subjects.jmp data table and a completed Subject Data panel.
Figure 4.16 Subject Data Table and Completed Subject Data Outline
Select Data Table
Select or open the data table that contains the subject data. Select Other to open a file that is not already open.
Subject ID
Unique identifier for the subject.
By
Produces a separate report for each level of the By variable. If more than on By variable is assigned, a separate report is produced for each possible combination of the levels of the By variables.
Construct Model Effects
Add effects constructed from columns in the subject data table.
For information about the Construct Model Effects panel, see “Construct Model Effects” in Fitting Linear Models.