Fig. 3: Identifying patterns of modality discrepancy through clustering. | npj Digital Medicine

Fig. 3: Identifying patterns of modality discrepancy through clustering.

From: Trait-like nocturnal sleep behavior identified by combining wearable, phone-use, and self-report data

Fig. 3

a Clustering of discrepant nights based on bedtime, wake time, midsleep time and time in bed (TIB) obtained from each modality along with WASO from Oura (n = 1755, discrepancy >1 h). Warm colors indicate later sleep timings (bedtime, wake-time, and midsleep) and longer durations (TIB, WASO) normalized per sleep metric, and cool colors indicate earlier sleep timings and shorter durations. b Plots show the discrepancy patterns for the resulting clusters for bed and wake time metrics. Estimates from each modality were compared to the overall average of the three modalities. Each boxplot displays the mean (black square) and the median (centerline) discrepancy values. The upper and lower hinges correspond to the first and third quartiles, and the upper and lower whiskers extend to the furthest point within 1.5 times the interquartile range from the hinges. c Illustrative samples of daily phone tap activity with sleep timings from the three modalities for each high-discrepancy cluster. Cluster 1: delayed Oura wake times (26th/27th May). Cluster 2: earlier tappigraphy-based bedtime relative to Oura and EMA (7th June). Cluster 3: earlier tappigraphy-based wake time (13th June).

Back to article page