Hello Erbil,

First, if you have more than one variable that you are tracking pre- and post-treatment, then you should consider a multivariate comparison (all variables/measures/DVs at a time) rather than a univariate comparison (one DV at a time). The only exception would be if you were convinced that all measures were uncorrelated (which seldom happens).

From there, I see two basic options:

1. Treat the pre-measure data as covariate(s), and run a one-way ancova (or mancova) on the post-treatment scores.

2. Run a one-between (groups), one-within (occasion: pre- or post-treatment) anova (or manova). If you choose this option, the test of interest is whether there is a group x occasion interaction, suggesting that the pre-post differences are unequal across groups.

There is, of course, the option of recasting the problem as a multiple regression model equivalent to either option above. But, the results of the tests of interest will be the same.

Either option requires assumptions about the data; most of these are easily checked.

Good luck with your work.

The choice between David Morse 's options 1 and 2 is often called Lord's Paradox. You might find the discussion on this useful for deciding between these choices (lots of papers exist, here is one I recently did, (onlinelibrary.wiley.com/doi/abs/10.1111/bjep.12300). If you need a copy not behind a paywall pm me.

In my opinion, you should run an ANCOVA, i.e. a regression:

post = pre + groups

There should not be an interaction between pre and group. See:

https://en.wikipedia.org/wiki/Analysis_of_covariance

Analysis of covariance provides an alternative solution where matching of groups is not possible before conducting experiments. In many experimental situations it is difficult to match the groups initially as it is not known that how many auxiliary variables are correlated with that of criterion variable and even if we decide to match the groups on few presumed correlated variables, it drastically reduces the size of the sample.

Analysis of covariance is essentially an extension of analysis of variance which allows for the correlation between initial and final scores. This analysis provides an adjustment in the final scores as per the differences in some initial variable. Analysis of covariance is popularly written as ANCOVA. Consider an ANCOVA experiment in which r treatments are compared with a scores in each group. If xij and yij represents the jth initial and final scorers in the group i respectively then total scores can be represented in the following tabular form.

Dear researcher, Erbil.

Here I have attached you're doubt related answer. Just read it and then get some ideas....

https://www.lehigh.edu/~wh02/ancova.html

Jos Feys , you are making the strong claim that the Erbil Murat Aydın is not interested in if the group effect is moderated by the baseline (pre) measure. Can you expand on this? I think this will be useful.

Daniel Wright , I think this link explains the assumptions for an ANCOVA quite weel:

https://en.wikipedia.org/wiki/Analysis_of_covariance

Thank you very much everyone for answers.

For example, I want to look at the effects of 2 different exercises on jump performance. Should I use ANCOVA or MANOVA, why just One-way ANOVA doesn't sufficient?