Hello,
I collected data to pilot test four different persuasive messages. I had each participant answer the same questions (Likert scale) after viewing each message. Therefore, to compare the difference between scores I ran a repeated measure ANOVA but split the data by stimulus to compare the scores of each message. However, since I split the data, my Mauchly's Test of Sphericity shows four lines, each with a different significance level. A line for each message.
When reporting the results, do I report each Mauchly's test?? Two are significant (assumption violated) and two are not (assumption not violated). Looking for some insight on how / what to report.
Thank you!
Hello Tess Buckley
I did not notice two points.
Abolfazl Ghoodjani, Tess Buckley said that each participant answered the same questions after viewing each message. As I understand that description, message is another repeated measures factor, not a between-Ss factor. But Tess will have to clarify.
Abolfazl Ghoodjani each participant saw all four messages and answered the same set of questions after exposure to each message. I wanted to look at the mean differences of each question between messages and within one message.
Hello Tess,
It sounds as if you have two within-subjects factors: (a) message (with 4 levels); and (b) occasion (pre/post exposure, thus 2 levels). The Mauchly test will only apply to the message factor, as it only comes into play when a repeated measures factor has more than two levels.
The four lines of Mauchly output in SPSS refer to different adjustment options for the presence of non-sphericity: (a) No adjustment (e.g., sphericity is assumed); (b) Greenhouse-Geisser adjustment; (c) Huynh-Feldt adjustment; and (d) Box's "lower bound" adjustment.
The first order of business is to look at the Mauchly test result itself ("Mauchly's Test of Sphericity"). If it is significant, then you should choose among the three adjustment options instead of using the "sphericity assumed" results. Note that what happens is that the degrees of freedom for the adjustment options are lower than those for the normal, sphericity assumed output results.
Both the Greenhouse-Geisser and Huynh-Feldt methods are less severe than the Box lower-bound method, so many folks would recommend you select one of these (again, assuming there was a significant Mauchly test result).
Report the line for each test that matches to the adjustment method you have selected.
Good luck with your work.
Bruce Weaver Yes, that is correct! Participants saw each message so believe the four messages are a repeated measure factor - not between.
Thank you for clarifying, Tess Buckley. But you also said that each participant answers the same questions (plural) after viewing each message. How many questions are there? Can they be summed or averaged to make scale? Or are they asking about completely different things? More info would be helpful.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA