*4.3. PSQI, Cognitive Scores, and SRSMs vs. Device Data*

Table 3 shows the results of a series of univariate linear models, each of which included either PSQI or cognitive score (morning, afternoon, and evening timepoints) as the dependent variable and the mean of the device metric per participant as the independent variable. The only statistically significant associations for PSQI at scale (i.e., significance threshold at alpha = 0.05) were for Oura's measurement of TSD and sleep efficiency (*p*< 0.05 for both). In both cases, an increase in TSD or sleep efficiency was associated with a significant decrease in PSQI score; since PSQI increases with poor sleep quality, these associations are in the expected direction (more sleep or more efficient sleep leads

to better or lower PSQI). Withings latency was statistically significant for afternoon cognitive scores and evening cognitive scores at *p* = 0.016 and *p* = 0.013, respectively. We did not find any significant associations between SRSMs and cognitive scores or overall PSQI.
