Abstract
Purpose: Most applications of conjoint analysis in health use choice tasks with 2 profiles, while marketing studies routinely use 3 or more. This study reports a randomized trial of paired and triplet profile choice formats experiments focused on hearing aids.
Method: Respondents with hearing loss were drawn from a nationally representative cohort complete identical surveys were randomized between choice tasks with 2 or 3 profiles. When they were offered, respondents also provided a full ranking of the 3 profiles cases. Baseline differences between the 2 groups were explored using ANOVA and C2 tests. The primary outcomes (i.e., the differences in estimated preference models) were explored using Wald and t tests and analysis of individual level models estimated by ordinary least squares.
Result: 500 respondents participated in the study, but 127 had no hearing loss, 28 had profound and 22 declined to participate and were excluded from analysis. of the remaining 323 participants, 146 individuals were randomized to the pairs and 177 to triplets, but the only significant difference between the groups was time to complete the survey (11.5 and 21 minutes, respectively). Pairs and triplets produced identical rankings of attribute importance but homogeneity was rejected (P = 0.0001). Pairs led to more variation, and were systematically biased toward the null, given a high proportion (32.2%) lexicographic respondents (i.e., respondents who did not trade across attributes), while all respondents in the triplet traded across attributes. The relative benefits of a full ranking also dominated pairs, but were not conceptually different form a single choice triplet.
Conclusion: The number of profiles in choice tasks affects the results of conjoint analysis studies. Here triplets are preferred to pairs as they avoid nontrading and allow for more accurate estimation of preferences models, but the benefits of requiring a full ranking of the 3 profiles are less clear.
Method: Respondents with hearing loss were drawn from a nationally representative cohort complete identical surveys were randomized between choice tasks with 2 or 3 profiles. When they were offered, respondents also provided a full ranking of the 3 profiles cases. Baseline differences between the 2 groups were explored using ANOVA and C2 tests. The primary outcomes (i.e., the differences in estimated preference models) were explored using Wald and t tests and analysis of individual level models estimated by ordinary least squares.
Result: 500 respondents participated in the study, but 127 had no hearing loss, 28 had profound and 22 declined to participate and were excluded from analysis. of the remaining 323 participants, 146 individuals were randomized to the pairs and 177 to triplets, but the only significant difference between the groups was time to complete the survey (11.5 and 21 minutes, respectively). Pairs and triplets produced identical rankings of attribute importance but homogeneity was rejected (P = 0.0001). Pairs led to more variation, and were systematically biased toward the null, given a high proportion (32.2%) lexicographic respondents (i.e., respondents who did not trade across attributes), while all respondents in the triplet traded across attributes. The relative benefits of a full ranking also dominated pairs, but were not conceptually different form a single choice triplet.
Conclusion: The number of profiles in choice tasks affects the results of conjoint analysis studies. Here triplets are preferred to pairs as they avoid nontrading and allow for more accurate estimation of preferences models, but the benefits of requiring a full ranking of the 3 profiles are less clear.
| Original language | English |
|---|---|
| Article number | BEC-12 |
| Pages (from-to) | E15-E15 |
| Number of pages | 1 |
| Journal | Medical decision making |
| Volume | 32 |
| Issue number | 2 |
| DOIs | |
| Publication status | Published - 2011 |
| Event | 33rd Annual Meeting of the Society for Medical Decision Making, SMDM 2011: From Evidence to Decision Making: Role of Behavioral Economics in Medicine - Chicago, United States Duration: 22 Oct 2011 → 26 Oct 2011 Conference number: 33 http://smdm.org/meeting/33rd-annual-north-american-meeting |