Two questions about a within-subject design

Dear all
I hope you are having a good week. I have an RR to a journal and one reviewer asks "Is there enough power to test three-way interactions? The authors should do, and then report, a power analysis.” BUT I don't know how to (and whether it's even possible) to calculate it in the context of my designs.
  • I have 2 selection studies (so not even experiments – people can browse a mock online news site and chose from within 12 articles that are pro-attitudinal, balanced or counter-attitudinal AND – within these – each is either attributed to a source of low/high credibility and either to a racial ingroup or outgroup), so it is ‘within-subjects’ so to speak (each person can choose as many or few articles as they want).
  • And there are no specific ‘factors’ BUT – again - the articles systematically vary the slant (pro/con/balanced), the credibility (low/high) and the race (white/black)
  • I use GLM with a 3-way interaction to test which ‘factor’ (slant/credibility/racial ingroup vs outgroup) most strongly predicts article selection.
  • AND in this context, I don't know how to do power analysis per reviewer’s comment. In the past, I have used some online calculators (e.g., but it doesn’t seem to be applicable…
AND one colleague told me that power analysis makes sense only with between-subject designs, in which you have experimental conditions but not for within subjects designs (where there are no conditions so you don’t need power analysis). Is this true? If yes, do you have any academic sources to back this up?

Second (related) question. The reviewer recommended to include socio-demographic controls BUT (again) the analyses rely on GLM models, which are within-subject, and so it is not methodologically ‘sound’ to include these controls because the within-subject nature of the GLM essentially includes an individual as his/her own control. Again, do you have any academic citations for this decision?

Please let me know. Your help is greatly appreciated.