Pennsylvania's Educational Quality Assessment Program provides each participating school with a building level report in which state percentiles are a prominent part. Multiple matrix sampling was being considered as a technique to reduce testing time. However, there was great concern that the error associated with estimating the school mean might lead to markedly different percentiles than obtained by census testing. Generally favorable results are reported from a post mortem simulation of multiple matrix sampling for a 2 to 6 subtest/subgroup sampling plan involving data obtained from over 30,000 students in 500 elementary schools. (Author/BW)