For each grade within each sampled school, a sample of students was selected from a listing of the students. Every student had an equal chance of selection. The sampling process was systematic (e.g., if the sampling rate was 1/2, then the sample was the first, third, fifth, seventh student, etc., on the ordered list, or the second, fourth, sixth student, etc., on the ordered list, determined at random). The sample size was set based on the sampled number of hits Hjs for the school. (In Rhode Island and Vermont, a reduced sample option was implemented at the request of the states. The number of hits for sampled schools in both states was reduced to 2 for grade 4 and to 3 for grade 8.) The preliminary sample size was 60*Hjs (60, 120, 180, . . . .). The "almost all" function was used to increase the sample size if the sample size was within a factor 60/65 of the total student enrollment for the grade, as was done for the school sample selection. For example, if Hjs was equal to 1 and the student enrollment was 61 through 65, then the sample size was increased to equal the student enrollment. In addition, fourth-grade schools were offered the option of taking all students if Hjs was equal to 1, and the school’s enrollment was 66 through 120. If the school agreed, all students were taken. If the school said no, a sample of size 60 was drawn. The great majority of grade 4 schools accepted this option. This option increased the grade 4 sample size in many states well beyond the 6,300 target.
Students then were randomly assigned to an operational reading or writing assessment, with one-half probability of each. This assignment was implemented by spiraling, in which the booklets given to sampled students were drawn from booklet packets containing one-half reading assessments and one-half writing assessments, in a randomized order.