|
Abstract:
|
For discrete choice experiments to be informative, respondents must actively weigh the tradeoffs between the alternatives in each choice task. Noting that respondent engagement is likely to vary across tasks, this project offers a practical solution for modeling respondent engagement at the task level. The proposed hierarchical Bayesian model uses task-level tracking data to allow for varying levels of engagement via task-level heterogeneity in the error term. Tasks wherein respondents are more engaged are weighted more heavily in the resulting likelihood, improving ?t and yielding better predictive results.
|
