Table 3. Modelling confidence ratings.

Estimated fixed-effect coefficients from generalized linear mixed-effect models.

		GLME
Experiment 1	Fixed-Effect	REDUCED 1	REDUCED 2	FULL
	Intercept (β0)	0.52±0.04 t5079 = 14.46; P = 1.90×10−46	0.72±0.02 t5124 = 39.00; P = 1.61×10−291	0.53±0.04 t5078 = 14.55; P = 4.92×10−47
	Choice difficulty (βΔQ)	0.33±0.06 t5079 = 5.77; P = 8.43×10−9	0.47±0.07 t5124 = 6.51; P = 8.18×10−11-	0.30±0.05 t5078 = 5.96; P = 2.73×10−9
	Preceding confidence (βc1)	0.28±0.04 t5079 = 7.60; P = 3.62×10−14	-	0.28±0.03 t5078 = 7.39; P = 1.67×10−13
	Context value (βV)	-	0.45±0.14 t5124 = 3.16; P = 1.58×10−3	0.47±0.14 t5078 = 3.21; P = 1.35×10−3
		GLME
Experiment 2	Fixed-Effect	REDUCED 1	REDUCED 2	FULL
	Intercept (β0)	0.53±0.03 t5145 = 17.57; P = 3.77×10−67	0.75±0.02 t5145 = 44.91; P = 0	0.53±0.03 t5144 = 17.12; P = 5.94×10−64
	Choice difficulty (βΔQ)	0.18±0.02 t5145 = 6.33; P = 2.63×10−10	0.25±0.04 t5145 = 6.51; P = 8.26×10−11	0.17±0.03 t5144 = 5.90; P = 3.85×10−9
	Preceding confidence (βc1)	0.29±0.04 t5145 = 7.01; P = 2.75×10−12	-	0.30±0.04 t5144 = 7.48; P = 8.54×10−14
	Context value (βV)	-	0.17±0.7 t5145 = 2.52; P = 1.18×10−2	0.16±0.06 t5144 = 2.51; P = 1.19×10−2

Note that the number of degrees-of-freedom differs between REDUCED GLME 1 and 2 in Experiment 1, because some participants failed to answer within the allocated time, causing missed observations. This has a lower impact on the number of usable observations in the REDUCED GLME 2 because this model does not make use of “preceding confidence” (which are missing observations–in addition to the missed trials- in the REDUCED GLME 2 and FULL FLME).