Could you possibly provide some more information?

What are these variables?

I assume you have 3 variables, 2 dependent and one 1 independed for two sets (pre test and post test)?

Are you sure its 2 depended and 1 independed? Usually is one depended (Y) and the others are independed (X1, X2,...Xn)

Then, what is your research question? Do you want to find differences in the variables pre and post test? Do you want to identify conguence between your variables?

It depends on the type of variables you have (nominal, ordinal, interval). Nevertheless, if you have two dependent variables (not related), a classic hypothesis testing approach should be advisable (1 test per dependent variable). Ie. t-test, Wilcoxon test, etc.

Before suggesting anything further, I would also advise what Panos Petsas said; we need more information about the research.

Panos Petsas Luis Medina Gual my study has two dependent variables, speaking scores and self-efficacy scores while the independent is a teaching method. I read books and they suggested me to use pretest posttest control group design if I do the experiment without randomizing the participants. The books suggested Ancova or Anova. But, in my case, I have two dependent variables. Thus, I am a bit confused on the best statistical analysis technique.

I Putu Indra Kusuma From my understanding, pre and post test is a variable as welll!

You want to test the differences pre and post test, right?

If one wants to test whether a discrette/continuous variable differs based on a categorical variable, ANOVA and ANCOVA are good suggestions.

For example, I would perform 2 ANOVA tests

1) Compare Speaking Scores pre and post test

2) Same for Self-eff. Scores, pre and post test

The last variable is teaching method. I do not know the methods you use, but lets assume it has three values (method 1, method 2, method 3). A categorical variable.

If you want to compare the time of test (pre or post), which is categorical with the teaching method (another categorical), you could use chi square test. I do not know if this has any meaning to compare though.

Laslty, since the two score variables are depended on the teaching method, you could use anova for them as well!

Panos Petsas Thanks for the suggestion. But, Anova and ancova are for one dependent variable. not for two dependent variables. Maybe Manova or Mancova would fit with it. But, I just do not know which one and why.

I Putu Indra Kusuma I understand, you want to examine them together!

Yes, these two tests are good.

(1) MANOVA is the Multivariate version of ANOVA, and it will test whether the multivariate means (your 2 scores) are different pre and post test.

My understanding is that your independed variable now is "pre or post test", right?

And the two scores are the depended.

(2) You could aslo check for MANCOVA, as it will let you blend other independed variables, like the teaching method. It works for additional categorical variables, like teaching method, and it will help you identify the effect of this variable as well!

Just note that these are parametric tests, that require some criteria to be met to be used efficiently (i.e., they assume normal distribution of your data).

I have seen them being used without validating these assumptions, but It might be better to look for some non-parametric equivalent tests.

For example, ANOVA has Kruskall-Wallis test, which is non-parametric and requires no assumption for data distributions!

I Putu Indra Kusuma

I think that the concept of MANCOVA suits you better:

1) It will help you deal with two depended variables (i.e., your scores).

2) It will also account for the different teaching methods (covariate).

3) And the most important, it will identify differences pre and post test!

I suggest that you take a look on this video guide. It will explain the benefits of MANCOVA over other tests.

https://www.youtube.com/watch?v=Xbv9SvWt1GI