RESUMO
The assessment of students in community settings faces unique difficulties. Since students are usually posted in small groups in different community settings and since the learning (largely) takes place outside the classroom, assessing student performance becomes an intrinsically complex endeavor. In this article, the proposition is made and tested that peers may be used to accurately assess particular aspects of performance, in particular those which need extensive and close observation. Examples are: Effort displayed while working in a community, quality of the interaction with that community, display of leadership, and subject-matter contributions.The Faculty of Medicine of the University of Gezira, a community-based medical school, uses peer evaluation to assess these attributes of its students. Thirty four students divided into three groups participated in the present study. Goals of the study were to appraise the reliability, validity and acceptability of an instrument designed to measure these characteristics. Reliability was estimated using generalizability theory. A high generalizability coefficient (G) was found when all items were taken into account. In this case, G equaled 0.97. Its sub-scales each had also fairly high G-coefficients: Effort: 0.89, Subject-matter Contribution: 0.91, Community Interaction: 0.89, and Leadership: 0,88 respectively. The validity of the instrument was studied using confirmatory factor analysis. The results suggested that the proposed model of peer evaluation is reasonably valid. Finally, the instrument turned out to be quite acceptable to students.