Then another type of reason for incomplete data is that the investigator planned it that way. This almost always occurs as a means of reducing the number of dominances that are directly observed. Several strategies for reducing the number of comparisons are possible. One is to systematically select a subset of all the possible pairs in advance and then develop as much of the order as is possible from the resulting dominances; sometimes it will even be complete. Another strategy is to present the stimuli in groups and have the respondents partially or completely rank-order them within the groups. The extreme possibility here, of course, is ranking the complete set.