Course · 2. Choice-based conjoint (CBC)
Your data
Chosŏn reads one row per product shown:
| resp | task | alt | brand | flavor | pack | price | chosen |
|---|---|---|---|---|---|---|---|
| 1 | 1 | 1 | Apex | Mango | 250ml can | 2.99 | 0 |
| 1 | 1 | 2 | Apex | Original | 250ml can | 1.99 | 0 |
| 1 | 1 | 3 | Nova | Berry | 250ml can | 1.49 | 1 |
| 1 | 1 | 4 | Apex | Original | 500ml can | 1.99 | 0 |
| 1 | 1 | 5 | 0 |
chosenis 1 for the product picked; every task has exactly one 1.- The none option is a row with every attribute empty.
- Respondents may answer different numbers of tasks, and tasks may show different numbers of products; nothing is thrown away. The data check says how much of each there is.
- Any other columns are taken as attributes. Chosŏn infers which are numeric and which look like price; you confirm or change that on the mapping screen.
Sawtooth .CHO files can be uploaded too. They hold numeric codes only, so add a design file with the level names in code order, or rename them after upload. (Our .CHO reader has been checked on the published format but not yet on a real export.)
The data check
Before fitting, Chosŏn checks that each task has one choice, every level is shown, levels are reasonably balanced, and the study is large enough. Errors block the fit; warnings explain weak spots.
The reference level
The first level of each attribute is the reference that others are compared with in the model. It does not change the results (part-worths are shown centred on zero), but willingness to pay is stated "versus the reference level", so pick one that makes sense, usually your own current product.
Try it in the tutorial: attributes and levels
Next: The model
Tutorial links open the example project in the app; it is free and needs a free account.