First of all, let’s define what we mean by pre-post data. The classic pre-post design consists in having measurements for an outcome variable in two time points, usually before and after an intervention or treatment. It is common two have two groups, one which will receive the full intervention and a second one, called the control group, which will receive a partial intervention or none. The interest of the researcher is to determine whether the intervention was effective in improving the levels of the outcome variable.
There are two different approaches to address this type of pre-post design: ANCOVA or repeated measures ANOVA. Either approach will work well in the situation described. The ANCOVA approach considers the post-intervention measure as the dependent variable and treats the pre-intervention measure as a covariate. The model looks for differences between the study groups after adjusting for any potential variation in the pre-intervention values between the two groups. The second approach consists of a classic repeated measures design with time as the within-subjects factor and study group as the between-subjects factor. In order to measure whether there is a mean change in the outcome study from before to after the intervention we will need to add the interaction term group*time.
We have seen that when there are two time points measurements both approaches, the ANCOVA and the repeated measures ANOVA are appropriate. However, there is a slight difference between them. The first one focuses on the differences between the post-intervention means, that is, which group has the higher mean after treatment, while the latter focuses on the mean change or gain.
What approach do we use when there is more than 2 time points measures? It is very common to measure the outcome variables more than one time after the intervention, that is to have one pre-test measurement and more than one post-test measurement. In this case you could either use a repeated measures ANOVA with all pre and post measurement included as the repeated measures or use a repeated measures ANCOVA with the pre-intervention measurement as covariate and all post-test measurements as repeated measure. In the latter, avoid using the pre-test score both as covariate and as the first measure in the repeated measures factor. This would lead to a bias in the results since you would be removing the intra-subject variation twice.
Finally, there is another technique, which you can always use when you have a design with repeated measures, linear mixed models. The main advantage of linear mixed models is that it allows to have different number of measurements for each individual. This is very common in social sciences where some subjects are sometimes lost to follow-up. While in ANOVA/ANCOVA you would have to disregard from the analysis all observations which are not complete, the advantage of mixed models is that it uses all the information for all subjects.
