I am running a study for my final year of a psychology degree. I've been having a hard time getting anything useful out of my supervisor, so I thought I'd turn to you very lovely and smart people.

My project is as follows:

  • I have two groups that receive a different kind of a meditation intervention. There is no control group in my study.
  • My outcome variables are measured for each participant before and after the intervention to see what impact the intervention had on them.
  • Outcome measures are tested with Likert-style questionnaires with responses ranging from 1 (least likely) to 6 (most likely).
  • I want to then see which type of intervention is better, hence also compare the two intervention groups to each other.

My questions are as follows:

  • What is the best way to describe the design I am using in this study? My main analysis is within subjects, as I am looking at each participant before and after the intervention. However, I will then compare the two groups with each other to see which intervention is better, hence introducing a between subjects component to it as well.
  • What would be the best kind of analysis for comparing differences in performance before and after the intervention considering that the responses are not of actual ratio measures, but of self-reported numbers on a scale?
    • After that, what would be the best way to compare the results of both of the groups to each other?

I know that this is a lot, but I am really struggling with this and cannot count on my supervisor's help as my whole university and all lecturers are striking at the moment so I am missing out on all of my contact hours and am constantly panicking over this.

Any help would be greatly appreciated!


2 Answers 2


An Independent T-Test is good for treatment comparisons. I assume the questions are the same for both groups? You might want to consider using subject matter experts to weight the questions as well (some questions may address more important benefits than others). You could have the participants assign weights as part of the survey, just make sure you assign the same mean weights to all the surveys before your analysis.

This site has a good overall process description for performing the test and reporting results...but chances are you don't have the software they use (still worth the read). https://statistics.laerd.com/spss-tutorials/independent-t-test-using-spss-statistics.php

With the Analysis Toolpak add-in turned on, you can perform the analysis in Excel. These instructions look pretty good: https://www.rwu.edu/sites/default/files/downloads/fcas/mns/running_a_t-test_in_excel.pdf

Good luck!

  • $\begingroup$ Hi Costgeek, sorry for the late reply as I have only seen your response. Your answer helps a lot and things make a bit more sense now! Thank a lot! $\endgroup$
    – Eileen
    Apr 18, 2018 at 16:52

It is a mistake to think of this as a within-subject analysis. Assuming you randomized to get two independent groups of subjects who received each of two interventions, what you have is a need for analysis of covariance. Here you would use the proportional odds ordinal logistic model on the final ordinal-scale measurements, adjusted for the baseline values. To flexibly adjust for baseline you might use quadratic functions of the scales, or just treat them as polytomous with 5 indicator variables. Do not compute any within-subject differences.


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.