Count-based analysis with bootstrap confidence intervals is the recommended default for most MaxDiff studies. Use the Windows desktop version from GitHub Releases when you want a local Windows app, or when you specifically need Hierarchical Bayes (HB), individual-level utilities, or larger model-based follow-up analysis.
Required format: Response ID | Attribute1 | Attribute2 | Attribute3 | ... | Most | Least
ā¢ Response ID: Participant identifier (can repeat for multiple tasks)
ā¢ Attribute columns: Items shown in each choice task (minimum 3)
ā¢ Most/Best: The item selected as most preferred
ā¢ Least/Worst: The item selected as least preferred
Segment file format: Response ID | SegmentColumn1 | SegmentColumn2 | ...
ā¢ One row per respondent with Response ID matching the MaxDiff data
Analysis method: Count-based analysis with optional bootstrap confidence intervals is the recommended default for most MaxDiff studies. For standard balanced designs, count-based scores usually give the same practical rankings with greater transparency. Hierarchical Bayes is only necessary when you need individual-level utilities for follow-up analyses like latent class segmentation, TURF optimization, or personalization. HB is available in the Windows desktop version, not in this browser version.