'The usefulness of Bayesian optimal designs for discrete choice experiments' by R. Kessels, B. Jones, P. Goos and M. Vandebroek

摘要

'Experimental design is perhaps the most important aspect of Discrete choice experiments (DCEs) as this determines what model(s) can be estimated with what levels of precision'. Never have truer words been so misunderstood, not by Kessels et al. but rather by many of those who do not fully understand or appreciate the message and meaning of their work. For those of us researching in this area, we would like to assume that the above sentiment is indeed the case and that designs do matter. Fortunately for us, the theory quite clearly and definitively supports that this is the case and happily, much empirical evidence also exists that suggests that the theory does indeed translate beyond the realm of simulated data. As both Louviere and Lanscar and Kessels et al. rightfully point out, experimental designs are generated under certain assumptions and depending upon what assumptions are made, the observed outcomes from the experiment may be influenced in very much unexpected ways. Indeed, one such way is the precision of the resulting parameter estimates obtained from the designs used in practice.