Abstract
We propose a simple strategy to construct D-, A-, G- and V-optimal two-level designs for rating-based conjoint studies with large numbers of attributes. In order to simplify the rating task, the designs hold one or more attributes at a constant level in each profile set. Our approach combines orthogonal designs and binary incomplete block designs with equal replication. The designs are variance-balanced meaning that they yield an equal amount of information on each of the part-worths.
Original language | English |
---|---|
Pages (from-to) | 3035-3046 |
Number of pages | 12 |
Journal | Journal of Statistical Planning and Inference |
Volume | 140 |
Issue number | 11 |
DOIs | |
Publication status | Published - Nov 2010 |
Externally published | Yes |
Keywords
- Two-level conjoint design
- D-, A-, G- and V-optimality
- Comparison depth
- Orthogonal designs
- Binary incomplete block designs
- Balanced and partially balanced incomplete block designs