A COMPARISON OF IMPUTATION TECHNIQUES FOR INTERNAL PREFERENCE MAPPING, USING MONTE-CARLO SIMULATION

The usual algorithm for internal preference mapping requires a complet e set of observations, meaning the technique cannot be used to analyse trials based on incomplete block designs. A simulation study was carr ied out to compare techniques for imputing missing values under variou s conditions. Sets of simulated preference data with different charact eristics were constructed. Monte Carlo simulation was used to create m issing observations in these sets; the imputation techniques were appl ied to the data; and the results of preference mapping based on the im puted data compared to those from the complete data set. Convergence p roblems were found with two techniques. Analysis of variance revealed that effects on performance were dominated by the proportion of data m issing, the level of noise in the data, and the size of the data set. Differences in performance among the three convergent imputation techn iques were small; mean substitution is recommended, as it performed as well as more complex iterative techniques. The results were broadly c onfirmed by a similar study on a genuine set of preference data.