Hi all,
I’m interested in finding out whether the preferences for visual cues (i.e. Y & Z) differ between two groups of subjects (i.e. A and B) across 5 different scenes (1,2,3,4,5). Each subject views each scene 2 times (yielding a total of 10 trials), and was supposed to provide a verbal response when they have reached a decision (i.e. trial duration was not standardised). Within each trial, each subject may view each visual cue more than once. In total, there were N = 760 fixations. Also, I have a small & symmetrical sample size of participants as well (14A & 12B)
After some reading, I think I’m supposed to use a Multilevel Logistic Regression. I’m aware of the violation of assumption on independence of observations – but this happens with eye tracking studies.
That said, I’m currently using SPSS. There doesn’t seem to be a button or place for me to account for the repeated measures element within the analysis GUI.
(1) Can I use a Generalised Estimating Equation (GEE)? If no, why?
(2) How should I organise my data so I can account for the within trial and within person measurements ? Should I just add another column Trial No.?
Currently, I’ve organised my data in long format with the following
(1) Dependent Variable = Visual Cues (Y & Z)
(2) Predictor Variables
(1) Participant Group (A & B),
(2) Scene (1,2,3,4 & 5)
If anyone can assist me in answering the above queries so I can ease into the analysis, that will be great!! Many thanks in advance!
i)You can use GEE too. (I hope GEE in SPSS supports this)
ii) Yes, your data should be in long format. You need indicators for visual cue, subject, time,scene and group (and others if you have).
Two references that might be helpful are:
https://www.stat.ncsu.edu/people/davidian/courses/st790/notes.html
HTH, David
Both methods should work. Normally, I would recommend multilevel analysis, because it allows inference on participant level, whereas GEE only does population-level effects. However, with the low number of repetitions and a binary response (which carries the smallest possible amount of information), there is no chance to reach reasonable levels of certainty on the participant level.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS