Im running a study that tests whether an intervention is effective in reducing depression levels, increasing well-being, optimism and self-esteem and increasing positive affect. My DV's are: Depression, Well-being, Optimism, Self-Esteem, Positive/Negative affect. my iv is the intervention itself.

There are 62 participants in the Pre-assessment condition and 70 participants in the post-assessment condition however its a within sample.

which statistical analysis should i use? A T-test or Anova?


1 Answer 1


With this data alone you will be unable to do any meaningful analysis of whether the intervention is effective.

You would need something that would tell you what would have happened, if the intervention had not been given. Patients will typically improve even in the absence of an intervention due to regression to the mean (especially if a certain patient state was required for entry), the placebo effect, the Hawthorne effect, more attention being paid to them as part of a study and so on.

Ideas for making some valid conclusions could be using historical controls (e.g. using the robust meta-analytic predictive approach or power prior approaches), concurrent controls (e.g. using propensity score methods, especially if you compare versus another intervention so that both interventions get a placebo effect) and other methods for observational studies.

  • $\begingroup$ the specific study was conducted on healthy students, it was basically a way to enhance well-being through a positive psychology intervention. There were controls but my analysis is only on the intervention (treatment group) $\endgroup$
    – Elena Mic
    Dec 30, 2018 at 14:11
  • $\begingroup$ An analysis in the intervention group only cannot tell you whether the intervention was effective. $\endgroup$
    – Björn
    Dec 30, 2018 at 17:26
  • $\begingroup$ Can you estimate the values of the DVs in the general (I'm assuming student) population? $\endgroup$
    – GuillaumeL
    Dec 30, 2018 at 18:39
  • $\begingroup$ @GuillaumeLoignon I guess if it is a representative sample of what happens in the population of interest, then I guess so. $\endgroup$
    – Björn
    Dec 30, 2018 at 18:54
  • $\begingroup$ Yes the sample is students. I now have a within sample and i applied paired t-test to the 40 participants i have in the intervention to asses the differences in the dv's before and after the intervention $\endgroup$
    – Elena Mic
    Dec 30, 2018 at 21:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.