6533b7d8fe1ef96bd126b536

RESEARCH PRODUCT

Analysis of Educational Frequency Data from a Complex Sample Survey

subject

description

Abstract Some recent methods are presented for analyzing categorial data from complex surveys involving clustering familiar in educational research where e.g. teaching groups are used as sample clusters. The methods are introduced through a discussion of the test of independence on a two‐way table and the analysis of a two‐way table using logistic regression models. The analyses are illustrated using data from the First National Assessment of the Finnish Comprehensive School 1979. The primary focus of the paper is on the methods that provide first‐order corrections to standard multinomial‐based chi‐square tests by taking account of survey design effects. Both first‐ and second‐order corrections take account at the sampling design in the sense that the actual type I error rates of the tests are closer to the nominal levels. In the case of a stratified cluster sampling as here, these corrections have a moderate effect to our statistical conclusions.