Dear Leard team,
I am currently experiencing some issues with my Master thesis data analysis and I would appreciate your help!!! First I will explain you my study design and then I will explain you my confusion. Hopefully you can help me out!
In my study, I would like to assess how my IV "sleep quality" affects my DV "aggression" on two different time points. We measured self-rated aggression on two time points, sleep quality only once. My supervisor insisted that I should use one-way repeated measures ANOVA to analyze the data, which I did. I used time as within-subject factor, sleep quality as a covariate and aggression as my DV.
Now I experience difficulties interpreting my results. For example, my analysis found the interaction between time and sleep quality to be significant. The main effect of sleep quality was also significant but not the main effect for time. So what do these results tell me? I don't know whether you can respond to this, but I really need an expert haha.
Kind regards,
Kristina
I suggest that you plot means versus trials keeping track of group. Connect the points by group(a profile). Because the interaction is significant, the curves should cross. Look at the order of means before and after the cross. This should aid in interpretation. This is often called a profile analysis. Examples are common in textbooks. For example see p. 458 of the reference below.
https://www.amazon.com/Multivariate-Applications-Education-Psychology-statistics/dp/0818500964/ref=sr_1_1?ie=UTF8&qid=1493220461&sr=8-1&keywords=Neil+H.+Timm
As David suggests you need to plot and interpret the pattern of means for the interaction. Main effects are not so interesting when an interactions present. An interaction means that the effect of sleep quality on aggression differs in magnitude or direction (or both) between the two time points.
The main effects are not so interesting because they are the average effect of sleep quality (for the two time points) and the average effect for time (over sleep quality). Having found an interaction you know that the these average effects might be misleading - hence look at the details of the pattern.
Dear Kristina,
In order to simplify your results, I would suggest - first of all - to calculate two additional variables: 1) difference between two time points of aggression, and 2) mean of two time points of aggression.
I. The interaction you are talking about ("sleep x time") means that sleep quality is linked to the difference between two time points of aggression. In the other words, sleep quality differentiates the changes in aggression pattern (please check: the linear regression between sleep quality and this difference should result with the same F-statistics and p-value).
II. The significant main effect of sleep quality means that there is a link between sleep quality and mean value of two time points of aggression (please check again the linear regression).
I hope these "supportive" analyses will work in your dataset. Please, give a feedback if it is ok or not ok.
Good luck!
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis