Let's say I have data on gene expression levels that were measured at multiple times during the day and night in various participants, e.g. in each of the 12 participants, the gene expression was measured every 2 hours, giving me 144 data points.
Does it make sense to calculate the z-score per participants and subsequently use a mixed effects model with participant ID as a random effect on the intercept to determine the effect of time (e.g. using a fit to a cosine curve) on the gene expression profile?
I don't think it is a reasonable approach because by calculating the z-score, the intercept becomes 0 in each participant, so including a participant as a random effect on the intercept won't add anything. Or might there be instances in which this could be useful/reasonable?
I'm asking this because someone asked me to do this kind of analysis.