0

For a meta-analysis I'd like to calculate and aggregate ES-values from studies reporting F-values for mixed model-ANOVA-s. Can you recommend any literature specifically on this topic?

+++

After posting the question I realized it is related to this one and this one. Only partial eta squared is easily calculated based on F-values and df. Olejnik & Algina (2003) discuss the comparability issues and offer a solution which, however, like eta squared requires more than is usually reported in papers to be calculated. Provided one calculated partial eta squared from studies with similar designs, comprising only manipulated factors, would it then be justified to aggregate these ES using "the usual" techniques, like sample size weighting?

Community
  • 1
skip
  • 130
  • 1
  • 2
  • 10
  • 1
    Check out Mike Lawrence's [answer](http://stats.stackexchange.com/a/6829/26743) to the linked question -- he also references a paper that deals with cross-study comparisons. – Livius Jun 23 '14 at 13:22
  • Thank you for your comment. If I read Olejnik & Algina (2003) correctly, for study designs comprising only manipulated factors partial eta squared is an appropriate and comparable ES measure. The next question would be how to aggregate partial eta squared-values from different studies. – skip Jun 23 '14 at 17:32
  • @Livius: you might consider reposting your comment as an answer, such that I could accept it (you offered an answer to the original question which was helpful). In this regard, see the CrossValidated_meta thread on comments that are actual answers ;-) – skip Aug 02 '14 at 19:58

1 Answers1

1

Check out Mike Lawrence's answer to the linked question -- he also references a paper that deals with cross-study comparisons.

Community
  • 1
Livius
  • 2,066
  • 14
  • 17