The Effect of Deleting Anchor on the Classification of Examinees

The presence of outlying anchor items is an issue faced by many testing agencies. This study examines the effect of removing or retaining one aberrant anchor item. The degree of aberrancy was manipulated as well as the ability distribution of examinees, and four IRT scaling methods were investigated (Mean-sigma, mean-mean, Stocking & Lord, and Haebara). The results indicate that the percent of correctly classified students was not affected by either retaining or removing the aberrant item, although the over- and under- classification of examinees was. There was no difference among the methods.

The Effect of Classmate Characteristics on Individual Outcomes: Evidence from the Add Health

We use data from the National Longitudinal Study of Adolescent Health (Add Health) to examine the effects of classmate characteristics on economic and social outcomes of students. The unique structure of the Add Health allows us to estimate these effects using comparisons across cohorts within schools, and to examine a wider range of outcomes than other studies that have used this identification strategy. This strategy yields variation in cohort composition that is uncorrelated with student observables suggesting that our estimates are not biased by the selection of students into schools or grades based on classmate characteristics. We find that increases in the percent of classmates whose mother is college educated has significant, desirable effects on educational attainment and substance use. We do not find much evidence that the percent of classmates who are black or Hispanic has significant effects on individual outcomes, on average. Additional analyses suggest, however, that an increase in the percent black or Hispanic may increase dropout rates among black students and post-high school idleness among males.