Table 4. Log hazard ratio estimate (standard error) per 100 g usual intake of each of the three vegetable subgroups, calibrated with each of the three forms of regression calibration models in their reduced and standard forms.

		Reduced form				Standard form
VegetableSubgroups	Calibration methods	(s.ea; s.eb)		s.e ratioc		(s.ea; s.eb)		s.eratioc
Leafy	Naïve method		−0.144 (0.027)				−0.144 (0.027)
	One-part linear calibration		−0.480 (0.090;0.112)		1.24		−0.409 (0.083;0.127)		1.53
	Two-part (untransformedDQ)		−0.395 (0.092;0.183)		1.99		−0.174 (0.089;0.278)		3.11
	Two-part (transformedDQ)		−0.509 (0.090;0.292)		3.24		−0.461 (0.047;0.160)		3.41
Fruiting	Naïve method		−0.094 (0.014)				−0.094 (0.014)
	One-part linear calibration		−0.125 (0.031;0.034)		1.11		−0.123 (0.031;0.034)		1.11
	Two-part(untransformedDQ)		−0.161 (0.030;0.034)		1.14		−0.109 (0.030;0.073)		2.42
	Two-part (transformedDQ)		−0.255 (0.037;0.108)		2.92		−0.228 (0.035;0.131)		3.74
Root	Naïve method		−0.160 (0.026)				−0.16 (0.026)
	One-part linear calibration		−0.342 (0.060;0.082)		1.36		−0.305 (0.054;0.077)		1.43
	Two-part(untransformedDQ)		−0.203 (0.088;0.219)		2.49		−0.107 (0.060;0.167)		2.78
	Two-part (transformedDQ)		−0.479 (0.070;0.214)		3.06		−0.265 (0.056;0.181)		3.23

s.e^a is the standard error (×10⁻²) for that does not account for the uncertainty in the calibration; s.e^b is the standard error (×10⁻²) that accounts for the uncertainty in the calibration; s.e ratio^c is the ratio of s.e^b to s.e^a.