*4.4. Cognitive Scores vs. Participant Summary Data*

Table 4 shows the results of univariate linear models that regressed cognitive score on participant summary features. With regards to the morning cognitive score, there was a significant association with the SF-36 sub-category of physical functioning (*p* = 0.014); however, further analysis revealed that this was due to the presence of an outlier with very low physical functioning as well as a low cognitive score, and exclusion of this individual removed the significant association. The SF-36 sub-category of emotional well-being was trending towards significance (*p* = 0.078) with cognitive score that appears robust to the removal of individual data points. None of the other summary features were significantly associated with the morning cognitive score. Several other features were statistically significant (*p* < 0.05) across two or more cognitive score timepoints, and also of note is the consensus of significance of features across afternoon and evening cognitive score timepoints. (Table 4).

**Table 3.** Results of multiple univariate linear models for PSQI (left) and cognitive scores across all timepoints (right). For the PSQI-related models, the independent variables were the means of device data for each participant, and the dependent variable was PSQI. The higher the value is on the PSQI, the worse the sleep quality; thus, positive correlations suggest relation to poorer sleep quality. For the cognitive score-related models, the independent variables were the means of device data for each participant, and the dependent variable were the cognitive scores. We show the *p* values of each univariate regression for cognitive score by timepoint. Please see Supplemental S6–S8 for more statistics related to these regressions. All units are in hours with the exception of wakeups (number of occurrences) and efficiency (a standardized metric).


Note: \* *p* < 0.1; \*\* *p* < 0.05; \*\*\* *p* < 0.01.


**Table 4.** In this collection of univariate linear models, the participants' summary data are the independent variables, and cognitive score is the dependent variable. We present the *p* values of each univariate regression for cognitive score by timepoint. Please see Supplemental S9–S11 for more statistics related to these regressions. These metrics all represent standardized scores.
