The Robustness of MDS Configurations in the Case of Incomplete Data

The authors examine the robustness of MDS configurations when incomplete rather than complete input data are used. Using two empirical studies, they show that robustness varies as the amount of incomplete data increases and that random methods of data deletion perform as well as cyclic designs. These findings provide empirical support for earlier Monte Carlo literature on the topic. The authors also show that individual characteristics of respondents, namely cognitive integration and imagery, influence the quality of configurations obtained with incomplete data.