TABLE 2.

Standard errors of prediction for point-normalized spectra

	RMS	Cross-valid lin1	Cross-valid lin2	Cross-valid quad3	Direct valid	Norm frequency	Frequencies in model	Model →% of the secondary structure
α-Helix	21.8	6.9	6.6	7.2	6.2	1641	1545–1565–1530	−37.082 + 223.655 × A1545 − 86.095 × A1565 − 82.481 × A1530
β-Sheet	17.6	6.6	6.6	6.6	6.1	1653	1638–1530–1656	−68.570 + 70.386 × A1638 − 21.433 × A1530 + 32.581 × A1656
β-Turns	4.2	3.4	3.7	3.6	3.1	1565	1685–1563–1687	102.655 + 31.310 × A1685 − 94.161 × A1563 − 23.584 × A1687
Random	11.4	8.6	10.4	10.4	8.4	1552	1679–1553–1687	−569.019 + 87.979 × A1679 + 593.514 × A1553 − 89.106 × A1687
310-Helix	3.2	2.8	3.0	2.8	2.6	1545	1546–1547– 1512	1147.907 − 2192.463 × A1546 + 1058.537 × A1547 − 15.803 × A1512
Random + turns + 310-helices	12.8	8.0	9.1	10.3	7.2	1543	1671–1587–1541	−252.957 + 47.236 × A1671 + 47.690 × A1587 + 242.377 × A1541

For legend, see Table 1. The spectra were normalized here so that the absorbance at the wavenumber indicated in the column Norm frequency is 1. The normalization frequencies were optimized for the linear model (Cross-valid lin¹) but not for the quadratic model (Cross-valid quad³). Note that the best values for the linear models for each structure between Tables 1 and 2 are in boldface type.