Latent variable modelling of discrete choice experiments

点击翻译按钮获取中文摘要

Discrete choice experiments and models are used to forecast consumer responses to changes in products policies and programs worldwide. Recent research suggests key model assumptions are violated because error variances covary with observed and unobserved factors. In order to address this, we will model systematic relationships between error variances and observed (eg, prices, survey length) and unobserved (eg, 'convenience', 'reputation') factors to improve model reliability and accuracy. This should lead to more accurate models/forecasts, benefitting business and government, which addresses the national priority of 'frontier technologies, promoting an innovative culture and economy'.