Table 1.
Parameter | Estimate | Unconditional s.e. | Confidence interval | Relative importance |
---|---|---|---|---|
Intercept 1|2 | −3.96 | 0.36 | (−4.67 to −3.25) | |
Intercept 2|3 | −3.70 | 0.36 | (−4.41 to −3.01) | |
Intercept 3|4 | −2.82 | 0.35 | (−3.51 to −2.15) | |
Intercept 4|5 | −1.79 | 0.34 | (−2.45 to −1.13) | |
Intercept 5|6 | −1.18 | 0.33 | (−1.83 to −0.53) | |
Intercept 6|7 | −0.18 | 0.33 | (−0.84 to 0.47) | |
Gender (female = 0) | 0.36 | 0.14 | (0.09 to 0.63) | 1.00 |
Incorrect (all correct = 0) | −1.95 | 0.28 | (−2.50 to −1.40) | 1.00 |
Age | −0.74 | 0.15 | (−1.03 to −0.46) | 1.00 |
Paranoia | −0.86 | 0.15 | (−1.16 to −0.57) | 1.00 |
Order (DG first = 0) | −0.02 | 0.08 | (−0.18 to 0.14) | 0.29 |
Punishment threshold was parameterised as a 7-level ordinal categorical variable, where lower levels indicate increased willingness to punish higher DG offers. For binary input variables, the reference category is given in parentheses. All continuous input variables were standardized and binary input variables were centred. Thus, estimates can be interpreted as being on the same scale. Importance is the probability that the term in question is a component of the true best model.