No. 1009

Title:

Some Statistical Properties of Japanese Self-administered Attitude Survey Samples Drawn with Schools or Classes as Units

Author(s):

Tsuchiya, Takahiro (The Institute of Statistical Mathematics)

Key words:

cluster sampling; homogeneity coefficient within clusters; self-administered survey; school-aged children.

Abstract:

    Cluster sampling with schools or classes as sampling units is often used in surveys that target school-aged children. This article examines the statistical properties of data from six Japanese self-administered surveys that employed cluster sampling. There were three important points suggested by the results. First, in order to obtain a nearly self-weighting sample via two-stage cluster sampling, it is proposed that schools should be sampled with probabilities proportional to the estimated number of classes in the case of elementary schools, and to the total number of enrolled students in the case of junior high schools. Second, variations between classes within a school are larger in lower grade. Hence, compared to one-stage cluster sampling, two-stage cluster sampling of the same sample size has more efficiency in higher grades than in lower grades. Third, responses to classroom-administered surveys show a higher degree of similarity than those administered at home. This tendency is substantial especially in young students of small classes in small rural schools.