In my study, participants were first randomly assigned to one of the four conditions (i.e., none, text, visual, video) for the 1st topic and answered the questions about psychological outcomes. Then, they were randomly assigned again to one of the four conditions (i.e., none, text, visual, video) for the other topic (2nd topic) and answered the questions about psychological outcomes.
Given that my team was interested in exploring if the same outcome was shown across different topics, which statistical approaches we should take for an analysis? I see one previous study conducted a multilevel modeling approach, but I am not quite sure if that approach is appropriate.
Many thanks for your time.
Assuming your outcome variable(s) are continuous (metrical), you could use (M)ANOVA to compare the means across conditions/groups (test for group mean differences).
Christian Geiser Thanks for your reply. The outcome variables are continuous, but wouldn’t simple ANOVA be problematic, given the participants’ exposure to two different topics (i.e., participants saw the message for the first topic and the other one for the second topic)? Having said that, the topic should be treated as a within-subjects factor, so not sure if ANOVA is still an appropriate method.
Hello Jiyoung,
I'm assuming that, by "randomly assigned," you mean that your participants could be assigned to the same condition on both topics. If not, you have a restriction (see third paragraph, below).
Consider an anova type design (or the analogous regression model) having the IVs of: (a) Topic 1 treatment (4 levels); (b) Topic 2 treatment (4 levels); a repeated measures dimension (questions for each topic), and possibly a multivariate DV (the psychological outcomes you mention). If you're only interested in interpreting one "outcome" at a time, then it's strictly univariate.
If you did not allow participants to be exposed to the same treatment for both topics, then you have an incomplete design with 12 possible sequences (e.g., "None-Visual," vs. "Video-Text," etc.). In this instance the sequence could be used as the IV; the RM and possibly multivariate DVs would also apply.
Good luck with your work.
David Morse Thanks for your detailed suggestions, David! You're correct in a sense that some participants could be assigned to the same condition on both topics (e.g., "text" for topic 1, and "text" for topic 2). Just another question: Is it allowed to use "repeated measurements" for the outcomes for each topic?
Hello Jiyoung Lee
If for the participants of the first topic, which is assigned condition A, the same condition A is selected in the second topic, then I think it will be a study of analysis of covariance.
But if you mean by randomly assigned that the participants in the first and second topics have different conditions, it is best to use a DOE model, especially factorial.
- Badges
- Science topic
- Similar topics
- Psychology