The statistical assumption of normality of difference scores is an important when analyzing continuous outcomes using a within-subjects design. When comparing continuous outcomes across time or within-subjects, the differences between observations of the outcome must be normally distributed.
In most instances, if the distribution of each observation of the outcome is normally distributed (skewness and kurtosis statistics above an absolute value of 2.0), then the assumption of normality of difference scores will be met. However, in order to be thorough and account for any inconsistencies in the data, subtract each observation from the other observations and run skewness and kurtosis statistics on the difference scores yielded to ensure that this assumption is met.
The assumption of normality of difference scores is assessed when using repeated-measures t-tests and repeated-measures ANOVA. Click on the buttons below to learn how to test for these assumptions in SPSS.
In most instances, if the distribution of each observation of the outcome is normally distributed (skewness and kurtosis statistics above an absolute value of 2.0), then the assumption of normality of difference scores will be met. However, in order to be thorough and account for any inconsistencies in the data, subtract each observation from the other observations and run skewness and kurtosis statistics on the difference scores yielded to ensure that this assumption is met.
The assumption of normality of difference scores is assessed when using repeated-measures t-tests and repeated-measures ANOVA. Click on the buttons below to learn how to test for these assumptions in SPSS.