I have survey data from a discrete choice experiment where each survey respondent was asked to complete 8 discrete choice questions (choose one) with each choice question having 4 unlabeled alternatives, where one alternative was a "pick none" option. There were 3 different survey versions and the 8 choice questions differed on each version, meaning there are 24 total choice situations across the 3 surveys.
I know choice data requires long formatting, but I want to make sure I am setting it up to correctly incorporate that each respondent completed 8 choice questions and that there were 3 different sets of choice questions a respondent might have seen.
This is an example of how my data is currently formatted. I have the survey version denoted, but label choice sets as 1-8 on all survey versions and alternatives 1-4 within all choice sets. Should the choice sets and alternatives be numbered differently? For example, choice sets numbered 1-24 (1-8 on survey 1, 9-16 on survey 2, and 17-24 on survey 3)? What about for alternatives... is 1-4 okay, or do they need to be numbered differently to indicate the alternatives they are choosing between differs between choice questions?
After correctly formatting the data, how can I make sure the clogit model is correctly interpreting this as one respondent making 8 different choices? If I group by ID it says there are multiple positive outcomes within a group since there are 8 ones in the "choice" column.
I know choice data requires long formatting, but I want to make sure I am setting it up to correctly incorporate that each respondent completed 8 choice questions and that there were 3 different sets of choice questions a respondent might have seen.
This is an example of how my data is currently formatted. I have the survey version denoted, but label choice sets as 1-8 on all survey versions and alternatives 1-4 within all choice sets. Should the choice sets and alternatives be numbered differently? For example, choice sets numbered 1-24 (1-8 on survey 1, 9-16 on survey 2, and 17-24 on survey 3)? What about for alternatives... is 1-4 okay, or do they need to be numbered differently to indicate the alternatives they are choosing between differs between choice questions?
| SurveyVersion | ID | ChoiceSet | Alternative | Choice | Price | Outages |
| 1 | 1 | 1 | 1 | 0 | 20 | 0 |
| 1 | 1 | 1 | 2 | 1 | 50 | 1 |
| 1 | 1 | 1 | 3 | 0 | 75 | 0 |
| 1 | 1 | 1 | 4 | 0 | 0 | 0 |
