The aim of this paper is to investigate two algorithms to rule out from the processing those responses that can be considered outliers in surveys based on L₄(2³)-type conjoint analysis. The evaluation of the algorithms is carried out checking the improvements in the precision of measurements. To accomplish it, a survey to assess the computer applications used in information literacy courses for female humanities students was conducted at a women’s junior college. Firstly, the raw data was processed and the precision values were computed. Then, the algorithms were used to establish the data set of outliers with a predefined maximum number of elements composing it. The values of precision of measurements were compared to evaluate the usefulness of these algorithms. Briefly, it was shown that the algorithms improved sensitively these values. Both algorithms performed in overall well compared to the raw data set; and superiority of one over the other depended on the cases. Finally, these algorithms are useful in cases in which a characterization of a group with a limited number of members is necessary to be done.

View full abstract