Item selection procedures designed for computerized adaptive need to accurately estimate every taker's trait level and,at the same effectively use all items in a bank.Empirical studies show that classical item selection procedures base on maximizing Fisher or other related information yielded highly varied item exposure rates.Some new criterions are proposed.The a-Stratified Multistage and OID methods procedure improves to a moderate extent the undesirable item exposure rates associated with the maximizing Fisher information criterion and keeps sufficient precision in 9 estimates.The maximizing Fisher information criterion will be compared with a-Stratified and OID in an empirical study using the mean squared errors inθand plots of item exposure rates associated with differentθdistributions. |