More Statistical Power Needed: The OHSU TREC 2002 Interactive Track Experiments
William Hersh, Susan Moy, Dale F. Kraemer, Lynetta Sacherek, Daniel Olson · 2002
Track experiments was to perform some preliminary experiments comparing searching on tablet devices versus ordinary personal computers. Unfortunately, the vendor who had promised the devices we planned to use was unable to deliver them in time for the experiments. We therefore shifted our experimental focus to assessing user factors found in previous experiments to be associated with success, with a particular desire to assess the role of spatial visualization. A variety of studies have demonstrated that spatial visualization is associated with successful computer use. Egan and Gomez have shown that spatial visualization is associated with two processes in text editing: finding the location of characters to be edited and generating a syntactically correct sequence of actions to complete the task [1]. Similarly, Vincente et al. have found that the ability to use a hierarchical file system is associated with spatial visualization as well as vocabulary skills [2]. In addition, Allen has demonstrated that this trait is associated with the appropriate selection of key words in searching [3]. We have previously found that the ability of medical and nurse practitioner students to answer clinical questions found spatial visualization to be highly predictive of success [4]. (Spatial visualization actually demonstrated multicollinearity with whether a searcher was a medical or nurse practitioner student, which may have been the actual predictive factor.) In previous TREC Interactive Track experiments, our results showed a trend towards spatial visualization being predictive of searching success in instance recall tasks, although they did not achieve statistical significance [5]. Methods Our methods employed the consensus approach agreed upon by track participants and posted on the track Web site