TABLE 3.
Unstandardized parameter estimates (and 95% Bayesian CI) of the unidimensional models.
| Target dimension | Non-target dimension | Absolute magnitude | |
| Threshold | 0.435 (0.311, 0.630) | 0.761 (0.616, 0.932) | 0.505 (0.396, 0.603) |
| RDHM | 0.771 (0.432, 1.05) | −0.770 (−1.08, −0.481) | 0.633 (0.424, 0.844) |
| RatioR effect | 1.65 (0.882, 2.41) | 2.54 (1.73, 3.31) | 1.83 (1.20, 2.47) |
| Format effect | 0.091 (−0.126, 0.319) | 0.043 (−0.193, 0.284) | 0.017 (−0.186, 0.221) |
| aRatioR → RDHM | 0.208 (−1.09, 1.53) | 0.636 (−0.556, 1.79) | 0.042 (−0.594, 0.660) |
| aFormat → RDHM | −0.132 (−0.497, 0.253) | 0.089 (−0.248, 0.440) | −0.306 (−0.507, −0.107) |
| aANS → RDHM | −0.07 (−1.52, 1.37) | 1.21 (−0.215, 0.261) | −0.584 (−2.12, 0.971) |
| aRDHM → Nline | 0.558 (−0.695, 1.79) | −0.978 (−2.21, 0.252) | 0.970 (−0.059, 1.99) |
| aRDHM → MPS | −0.02 (−0.904, 0.824) | 1.24 (0.387, 2.09) | −0.861 (−1.57, −0.096) |
Top and bottom panels refer to item-level and subject-level effects, respectively. Parameters in bold indicate that the 95% CI does not cross zero. Ratio R refers to the ratio of ratios. MPS, Math problem solving; NLine, Number line estimation skills. The threshold in the probit regression is estimated at the mean-centered ratio of ratios and subject-level variables. Contrast coding is used to code RDHM (i.e., −0.5 and + 0.5 for congruent and incongruent trials, respectively) and format (i.e., −0.5 and + 0.5 for continuous and discrete representations, respectively) so the threshold corresponds to the overall mean probit.
aIndicates a cross-level interaction.