Statistical inference in educational research

Summary

This article follows a recent proposal in Educational Research for the use of statistical tests in situations where the normal assumptions of random sampling or random assignment are not met. It supports such a proposal, but offers an argument for these procedures in terms of testing an observed pattern of association or difference against a model of random distribution of characteristics rather than as part of ordinary descriptive procedures. The value of statistical significance testing is also emphasized, as is the importance of designing research in a way that will meet the normal assumptions connected with inferential techniques.