Finding Unexpected Response Patterns in TGMD-2 Using Social Network Analysis

TGMD – 2 ( Ulrich, 2000) is a well-known standardized test to assess gross motor skills for children age 3 to 11. The test consists of 12 skills – six skills each for two subtests, locomotor and object control. It is assumed that students who have similar or same summary scores would respond in a similar manner across all items. In practice, subjects do not follow expected patterns. In aptitude assessment, those unexpected patterns have been used to identify invalid response patterns, such as cheating or guessing (Meijer & Sijtsma, 1995, 2001), and in personality domain, they were used for broader purposes such as detecting social desirability (Zickar & Drasgow, 1996) or person-fluctuation (Ferrando, 2004). These unexpected patterns, which may provide rich information for instruction or intervention, have been basically ignored in motor assessment. Purpose: To detect children who showed unexpected response patterns in TGMD-2 and describe characteristics of the patterns using social network analysis approach. Methods: The data from the TGMD-2 study (Ulrich, 2000) were used for the study. Analysis/Results: The binary raw data were converted into the dissimilarity measures, on which the residual matrix was calculated by subtracting observed measures by expected measures. According to residual differences between participants, group memberships were determined using Ucinet – social network analysis software (Borgatti, Everett, & Freeman, 2002). Descriptive statistics were then computed to compare group difference in their response patterns. About 20% of subjects (N = 210) were not classified. The rest of the children were classified into four groups according to both proportion scores and response patterns: High (% of subjects = 48%, mean score = .86), moderately high (14%, .76), low (11.5%, .53), very low (6%, .25). Age played a major role in determining these groups. Among the four groups, high and low ability group had subgroups which showed unexpected response patterns. Within the high group, the one small group (5%) showed exceptionally low passing score (.35) in all four subskills of overhand throw compared to the other major group (42%, .80) and also similar patterns in three difficult items related to arm movement subskills. Within the low ability group, two subgroups illustrated unexpected response patterns: One group showed unexpectedly higher proportion scores in locomotive items and another group higher in objected control skills. Conclusions:Social network analysis is a useful approach in identifying unexpected response patterns in motor assessment, which in turn can be used to help design appropriate instruction and assessment.