Table 5.
Model Constraint Testing
| Factor | Outcome | Value | df | p | |
|---|---|---|---|---|---|
| LANGUAGE | UNTIMED | 31.69 | 3 | .005 | |
| LANGUAGE | TIMED | 11.93 | 3 | .008 | |
| LANGUAGE | ALL | 33.99 | 6 | <.001 | |
| ATTENTION | UNTIMED | 7.10 | 4 | .131 | |
| ATTENTION | TIMED | 6.56 | 4 | .161 | |
| ATTENTION | ALL | 11.99 | 8 | .151 | |
| WM | ALL | 9.96 | 2 | .007 | |
| PS | ALL | 1.30 | 2 | .521 | |
| Factor | Outcome | Estimate | S.E. | t = Estimate / S.E. | p |
|
| |||||
| PA | UNTIMED | 10.97 | 3.88 | 2.830 | .005 |
| RAN | UNTIMED | -12.38 | 2.80 | -4.427 | .000 |
| Vocabulary | UNTIMED | 0.41 | 0.63 | 0.652 | .515 |
| VAS | UNTIMED | 0.44 | 2.42 | 0.180 | .857 |
| Visual Search | UNTIMED | 5.13 | 3.07 | 1.671 | .095 |
| CPT | UNTIMED | 3.33 | 2.09 | 1.593 | .111 |
| Behavioral Attention | UNTIMED | 1.09 | 1.85 | 0.592 | .554 |
| WM | UNTIMED | -11.08 | 3.80 | -2.919 | .004 |
| PA | TIMED | 3.69 | 2.74 | 1.346 | .178 |
| RAN | TIMED | -6.85 | 2.55 | -2.689 | .007 |
| Vocabulary | TIMED | 0.58 | 0.57 | 1.020 | .308 |
| VAS | TIMED | -0.14 | 2.28 | -0.062 | .951 |
| Visual Search | TIMED | 1.58 | 2.71 | 0.584 | .559 |
| CPT | TIMED | 4.07 | 1.94 | 2.094 | .036 |
| Behavioral Attention | TIMED | -1.98 | 1.70 | -1.167 | .243 |
| WM | TIMED | -7.17 | 3.06 | -2.341 | .019 |
| PS | READING | -2.74 | 2.46 | -1.111 | .266 |
| PS | MATH | -0.77 | 2.66 | -0.288 | .773 |
Note: UNTIMED = untimed achievement measures (KTEA-3 Letter Word Recognition and Math Computations); TIMED = timed achievement measures (KTEA-3 Word Reading Fluency and Math Fluency); ALL = both timed and untimed achievement measures; READ = reading achievement measures (KTEA-3 Letter Word Recognition and Word Reading Fluency); MATH = math achievement measures (KTEA-3 Math Computations and Math Fluency). Top half of Table: Wald tests constraining all indicators of a specific cognitive domain to be the same across outcomes; for example, the first line (LANGUAGE, UNTIMED) tests whether the four language measures collectively can be constrained to be equal across the two untimed achievement measures (they cannot without negatively effecting model fit). Bottom half of Table: individual model constraints evaluating whether pairs of estimates (one across two outcomes) are different; for example, the first line (PA, UNTIMED) tests whether the PA factor can be constrained to have equal loadings across the two untimed achievement measures (they cannot without negatively effecting model fit).