I have multiple IVs (4 personality traits, age, gender and frequency use of social media and my DV is whether participants contributed to the circulation of fake news online. I had 3 news posts in the questionnaire (1 real news story and 2 fake news stories).
The participants answered on a scale of 1-5 and for each of the 3 posts, they were asked 'are you likely to like or comment, are you likely to discuss with others online, are you likely to share.
I then combined the answers into 1 variable labelled 'engagement'. So I now have an 'engagement' variable for each of the 3 news posts. This means I now have 3 DV's.
I am really unsure which analysis to run for this data, as I can't find how to analyse the effect of the IV's on each DV and then see if there were differences between each DV.
Any help on this would be really appreciated as I'm at a loss. Thanks in advance!
Dear Sophie,
you might find this overview from UCLA helpful to choose a statistical test in different software for statistical analysis: https://stats.idre.ucla.edu/other/mult-pkg/whatstat/
Good luck!
Dear Sophie,
It depends on your design. Hierarchical Multiple Regression analysis or MANOVA can go. Thanks.
Dear Sophie,
I suppose I'd have to know more to be able to form a firm opinion but, given what you've outlined, it seems to me that you might want to consider a Canonical Correlation analytic approach. It will allow you to simultaneously consider relationships among multiple IV and DV variable sets. Tabachnick and Fidell have a really nice explanation of the analyses in a chapter in their book that you might consider. All the best. Bob
Hi Sophie, similar to Dr. Eklund it sounds like a CC analysis would be a good choice so you can analyze a set of IV and a set of DVs. Another option if you have theoretical or conceptual support for a mediator and/or moderator (age?) is a mediation analyses and you would have to find the right model after downloading an add on to SPSS from Preacher and Hayes. Good luck. Jeff
I will add sample size is always an issue in all research, and canonical correlation probably requires a larger sample than simpler statistical methods. The point is--it depends on many factors! You could run several regressions, but that has a downside as well.
- Badges
- Science topic
- Similar topics
- Comment