iPod Track Listening & Biological Samples— Within-Subjects Test/Regression? I am analyzing data in which participants listened to iPods and provided biological samples on the same days. The iPod Listening Intervention was 1 week. The biological sample collection was 4 days within the 1 week. There was no intervention/control group. All participants in study were given iPods and materials for biological sample collection. 
Research Question: Does listening to the iPod affect biological samples on days listened?
I’m looking to match dates of collection with listening dates to see if there is a difference between the biological samples (continuous variable) on days listened and days not listened. Importantly, some participants did not listen at all, so this may provide useful data. Some participants listened every day of the intervention while others only listened a few days. 
I am assuming that participants who listened to tracks on the days of collection will provide healthier bio samples whereas days not listening will not be as healthy. 
I’m very interested in this research question and was given an awesome dataset to work with, but don’t know exactly where to start. It seems like a within-subjects design but potentially could be measured by regression. 
(I have already run regressions examining total listening time with the biological sample, but am looking to answer a slightly different question). 
Thanks!
 A: A common approach to this would be a repeated measures ANOVA. This assumes that repeated measures within a given individual are more correlated than measures between individuals (conditional, of course, on whether they did or did not listen to music that day). 
This is similar to a random effects model with a "random intercept" that explains a summation of all their unmeasured individual level predictors of the outcome, so that measures in the same individual are effectively independent.  They are not identical but very close.
A critical assumption for these types of model is a lack of lagged effects. That means that when I listen to music, my "biological sample" changes until I stop. When I am measured the next day, not listening to music, my "biological sample" is typical of a music free day for any amount of time. A way of testing this assumption is creating and testing a lagged exposure as a fixed effect: did a participant listen to music on the previous day and did that affect their health today. This is called an AR-1 type autoregressive dependence. You can consider any number of lagged effects. It is a good exploratory analysis to enrich the findings. 
