Fig. 2: Cohort similarity and example feature separability. | npj Women's Health

Fig. 2: Cohort similarity and example feature separability.

From: General feature selection technique supporting sex-debiasing in chronic illness algorithms validated using wearable device data

Fig. 2

A Cohort filtering diagram. All individuals were first filtered by having at least 1 usable day of data as well as confirming one of CNC, H, D, or B. Then individuals were filtered as having at least 28 days of data. A subset was then selected for age over 40 years. Counts are shown at each step for the total cohort, females in blue, and males in red. B Features by Sex, Age, and Condition Dendrogram. Median sample of cohorts compared to each other with agglomerative clustering. Each color identifies a cluster that exceeds an average distance threshold from other clusters. Example features significant in exclusively females or males. Non-singleton clusters ordered top to bottom as 1, 2, and 3 and separated by black dashed lines. C Example feature, daily temperature kurtosis, with significant separability when considering condition or sex; however, when considering condition and sex, is only significant in females (blue) and not males (red). D Example feature, nightly pCorr respiratory rate (RR)/heart rate variability (RMSSD), that is statistically different in males but not females separated by the same condition. Two-sided, Mann-Whitney U, ***p < 0.001.

Back to article page