If there are two groups (IV: treatment / no treatment) with say 15 participants each, and for each participant the same observer has completed a questionnaire (DV: outcome measure, completed once for each participant by the same observer; interval scale), what kind of data analysis would be appropriate? I am interested if participants in the two groups differ on the outcome measure, which has been completed by the same observer for all participants.

For a dependent t-test, data structure would look like this

group 1 (control group): 2, 4, 3, 2, 2, 5, 3, 2, 3 etc.

group 2 (treatment group): 3, 2, 4, 4, 1, 2, 3, 1, 1 etc.

but I feel this would not make sense as it is not a pairing within the "participants" (for each participant, there is only one score); all data is provided by the same observer.

Also, as a second step: What approach could be used for two groups with 30 "participants" each where observer A contributes 15 values per group and observer B also 15 values per group?

  • $\begingroup$ You stress that both groups have same 'observer"? What do you mean by 'observer' and you do you mention there is only one? $\endgroup$
    – BruceET
    Jul 28, 2020 at 17:02

1 Answer 1


This kind of sounds like a repeated measures experiment, where a group undergoes testing, and then they repeat that same test under a single different condition (in your case a later time).

A little more info would be helpful, like what are your dependent and independent variables, are you testing the observer, or the participants? From my current understanding of your question, a paired samples T-test would probably be the way to go, if you are comparing means.

As for your second part, i would run a 2x2 anova to properly analyze that. you have group A doing task 1, and task 2. You then have group B doing task 1, and task 2. this would be best comapred using a repeated measures anova, with the post-hoc test of your choosing.

  • $\begingroup$ Thanks for your response. I have edited the original post and added some more information. I was thinking of a dependent t-test too, but I would not know how to enter the data in a meaningful way, as there are no paired datasets as there usually are with a dependent t-test. $\endgroup$
    – grey
    Jul 28, 2020 at 13:05
  • 1
    $\begingroup$ thank you for the clarity. I think the best way to go about this would be to use the 2 groups and compare with a t-test as you normally would. it doesn't really matter that there was 1 person observing (it might actually be better as it rules out a third variable problem), as long as his observations remained impartial throughout. so I would say to run the test as you would like any dependent t-test. I have also added a little more about your second step in my answer $\endgroup$
    – user292603
    Jul 28, 2020 at 13:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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