Abstract
Personality disorders (PDs) are most often evaluated on the basis of self - report, despite involving the way that one's behavior affects others. Nearly all studies of peer perceptions of PDs have relied on self - selected informants, which may result in low reliability and overly positive biases. Although obtaining information from large groups of unselected peers is preferable, it introduces complicating effects of group dynamics. In addition, in a large group of peers, not all raters will make equally valid ratings of all targets. The present study utilizes social network analysis to investigate ways of improving reliability and validity in peer ratings. Participants were 21 groups of peer raters from a military population (N:809) who acted as both targets and raters in a roundrobin design. Using the Peer Inventory for Personality Disorder, individuals identified other participants who exhibited traits of DSM - IV personality disorders. Participants also completed self - report versions of the same instrument. Mixed linear models were used to estimate variance in peer ratings due to network, rater, target, rater - target interaction, and self - report. Adjacency matrices were constructed based on participants' self - report of how well acquainted they were with one another. Social network analysis was then applied to find network characteristics of participants, and to identify a variety of cohesive subgroups within networks. Network characteristics were associated with both self -and peer - reported personality disorder traits. Consistent with DSM - IV descriptors, measures of centrality iv and degree connectivity were positively associated with narcissistic and histrionic PDs, and negatively associated with avoidant, schizoid, and schizotypal PDs. Peer ratings made within cohesive subgroups were larger, had higher self - peer agreement, and were more reliable, than did those made by raters who did not share a mutual subgroup with the target. Partitioning networks into two subgroups achieved improvements as large as, and more consistent than, identifying small tight - knit cohesive subgroups. Social network analysis is posited as a means of incorporating aspects of Kenny's (1994) Weighted - Average Model in a cruder, but more parsimonious, way. It is recommended that researchers investigating peer perceptions of normal and abnormal personality consider partitioning large groups into two cohesive subgroups, to maximize reliability and validity of ratings.
Note: Abstract extracted from PDF file via OCR
