For my study, I am using an fMRI dataset of people watching movies. I computed values for each word in the movie, and am creating X amount of bins (now 3, but might become more) to categorise those values (in the three bins now, I have low 0-30, medium 30-60 and high 60-100 values). I have 20 participants, and want to compare these bins (in their fMRI activity) to see if the activity in these binned categories are significantly different from each other.
Does anyone have an idea which statistical test would be the most appropriate to use? I was thinking of 3dANOVA, but I am not sure which one to pick, or if I could somehow solve this issue with 3dttest(++) (although this is likely not the best option, as I want to possibly expand the categories from 3 to 5).
Any help would be hugely welcome and appreciated!
Binning the data, in general, is not recommended. Instead, it is preferable to utilize the original data and evaluate the marginal effect.
Does anyone have an idea which statistical test would be the most appropriate to use? I was thinking of 3dANOVA, but I am not sure which one to pick, or if I could somehow solve this issue with 3dttest(++)
Could you clarify whether the analysis being referred to is conducted at the individual or population level? Providing additional details about the data structure would be helpful in further clarifying the context.
Thank you very much for your reply, excuse me for the lack of context.
I have am conducting a 3dDeconvolve at the individual level, and now want to look at the group level.
Regarding the data structure, I have the fMRI activity for the whole duration of the movies. I have files with all the words in the movies and their timing, duration, frequency and predictability. In this way, I can approximate the fMRI activity during the words.
I want to include age, gender, duration and word frequency as covariate variables. Furthermore, I was already considering to not create bins, but treat the data as continuous (so could be considered repeated measures at a specific context window, e.g. 50 words are a measurement of an optimal context window of 1, 47 words are a measurement of optimal context window of 2, etc. up to 100).
Considering this complex design, I think that 3dLME would make the most sense, but I am not sure how I could implement that. Is there any other information that would create more clarity?
I want to include age, gender, duration and word frequency as covariate variables.
I'm still having difficulty understanding your data structure. Is the context window or the number of words your variable of interest? It would be helpful if you could provide an example for one participant in a table format, as it would give me a better grasp of the basic idea.