ABSTRACT
This chapter considers adjustments to the degrees of freedom to compensate for correlated measurements over time. Researchers for some time has been concerned about the effect of correlation on analysis. If two measurements are perfectly correlated, they together represent only one piece of information, or one degree of freedom. The chapter also considers multivariate analysis relevant to repeated measures. Multivariate analysis seems quite appealing as it makes few assumptions about the form of correlations over time. These multivariate tests complement plots over time and univariate statistics for each time period and for polynomial contrasts over time. If the sphericity conditions hold, then contrasts over time behave as they would for split plot designs. Split plot analysis is often justified if correlations have a certain form. A more 'sophisticated' approach is possible by using an array of multivariate tests which allow for arbitrary correlation structure among and within subjects.
