Table 3.

Linear model M₁ full-cohort fit results for daytime SpO₂ (top row of coefficients) and nocturnal SpO₂ (bottom row of coefficients).

Linear regression results for proposed model (M1)
Model	R2	SEE	Constant	Age (years)	BMI (kg/m2)	Altitude (km)	Sex (m = 1, f = 0)	Asian race/ethnicity	Black race/ethnicity	Hispanic race/ethnicity	Other race/ethnicity
Full-cohort daytime SpO2	0.23	1.12	96.66 (96.62, 96.70) [p < 1e−10]	−0.0314 (−0.0327, −0.0300) [p < 1e−10]	−0.0457 (−0.0484, −0.0430) [p < 1e−10]	−0.8402 (−0.8845, −0.7959) [p < 1e−10]	−0.0490 (−0.0880, −0.0101) [p = 4.1e−04]	−0.1004 (−0.1735, −0.0273) [p = 1.2e−04]	−0.0648 (−0.1446, 0.0150) [p = 0.023]	−0.1334 (−0.1936, −0.0733) [p = 4.8e−10]	−0.0146 (−0.0986, 0.0694) [p = 0.63]
Full-cohort nocturnal SpO2	0.32	1.21	95.95 (95.91, 95.99) [p < 1e−10]	−0.0429 (−0.0444, −0.0415) [p < 1e−10]	−0.0562 (−0.0591, −0.0533) [p < 1e−10]	−1.1736 (−1.2214, −1.1257) [p < 1e−10]	−0.0038 (−0.0459, 0.0383) [p = 0.80]	0.0775 (−0.0015, 0.1564) [p = 0.0059]	−0.0068 (−0.0930, 0.0794) [p = 0.82]	0.0243 (−0.0406, 0.0893) [p = 0.29]	0.0596 (−0.0311, 0.1504) [p = 0.065]

Sex variables for the full subject cohort are encoded using a value of 1 for male subjects and 0 for female subjects. Race/ethnicity variables for the full cohort and both sexes encoded using a value of 1 for each listed race/ethnicity group, with White subjects encoded using all zeros. Values listed in parentheses represent 99.5% confidence intervals for the fitted model coefficients. Values in brackets are regression coefficient p values, corresponding to two-sided t tests under the null hypothesis that the coefficient is equal to zero.

R² coefficient of determination, SEE standard error of the estimate.