Sawtooth Software: The Survey Software of Choice

Nearly every CBC and MaxDiff project fielded today involves multiple versions (blocks) of the experimental design plan. If an unequal number of respondents completes each version, does this cause much practical consequence? Keith Chrzan investigates this question using carefully simulated respondent data (imitating patterns of preference from real CBC and MaxDiff studies) and concludes that even cases of extremely imbalanced incidence across versions poses no trouble for true parameter recovery. A caveat is that Keith’s simulations cannot account for potential psychological order and context biases, which the author discusses and invites other researchers to investigate.

  Download PDF    View in new window