Extended Data Fig. 3: Linear Regression and Double IRF prediction models with respect to HbO and HbR.
a) Maps of the weights, A (left) and B (right), in the Linear regression model for HbO (n = 8 subjects). b) HbO timing coefficients for Ca2+ and NE (tA, tB); each point corresponds to an average across runs for one subject (n = 8 subjects; mean ± SEM; tA = 0.39 ± 0.08, tB = 0.30 ± 0.07). c) Linear Regression model performance map quantified as the correlation between the experimental and predicted HbO (n = 8 subjects). d) Difference between HbO Linear Regression model performance and HbT Linear Regression model performance (n = 8 subjects). e-h) Same as (a-d) but for HbR. For (f), tA = 0.28 ± 0.08, tB = 0.52 ± 0.17. i, k, l, m, o, p) Same as (a, c, d, e, g, h) but for the Double IRF model. j, n) Estimated IRFCa2+ and IRFNE for the HbO and HbR model (n = 8 subjects; mean ± SEM). q) Side-by-side model comparison. For each subject and model, the model performance (r) was averaged across space. Each point represents one subject (*p < 0.01 two-sample, two-sided Kolmogorov-Smirnov Test; mean ± SEM; r = 0.32 ± 0.02, 0.29 ± 0.03, 0.39 ± 0.02, 0.60 ± 0.03, 0.59 ± 0.03, 0.69 ± 0.02, 0.67 ± 0.03, 0.67 ± 0.01, 0.65 ± 0.02; n = 16 subjects for Global, SSp, and Variant models; n = 8 subjects for Double IRF and Linear regression models). P-values are provided in figure source data.
