I have some timecourse data which plotted looks like the figure below. I want to better describe the difference between the two conditions.
My adviser advised :D me to use a linear model and observe a time effect (interesting!) a main condition effect (probably none) and a condition*time effect (most interesting).
Now, I am wondering what this information would look like - the time course isn't gradual, and I have run similar tests with R's lme4::lmer
and nlme:lme
and for continuous variables (in my case time) they just output one value as if the progression is linear. Obviously in my case it is not :-/
I have previously worked with such data and used PCA (which indeed, does not give me just-linear progressions) - would it be a good substitute for a linear model in my case? Also, the problem with PCA is that I have no idea how to reap the added information one would get from single participants (with mixed models, I can just add the participant IDs as random variables and feed it the raw data) - is there any way to pass the entire (un-meaned) per-participant data to, say Python's mdp.pca()
and get something reasonable out of that?
Participant
effect. Something like:lmer(PC1_scores ~ Var1 + Var2 + (1|Participant)
. (If I understand what you mean by your PCA problem.) $\endgroup$