Independent scores

In between-subjects designs (e.g., the independent t-test or one-way ANOVA), data from different participants should be independent meaning that the response of one participant does not influence the response of another participant. We violate this assumption in the case of nested data (e.g., when our sample consists of students in three different classrooms, it is likely that students within classrooms are more similar than we would expect otherwise).

In within-subjects designs (e.g., the dependent t-test or repeated measures ANOVA), we automatically violate the assumption because of course the scores of one participant in one condition will relate to their scores on another condition. However, their scores should still not influence any other participant’s response.

This is another assumption, like interval/ratio data, that we do not ever test but is a function of knowing our data.