Skip to main content
. Author manuscript; available in PMC: 2022 Jan 1.
Published in final edited form as: J Educ Psychol. 2020 Feb 27;113(1):86–103. doi: 10.1037/edu0000467

Table 3.

Main Effects Multi-Levels Model Results (n=391)

Model/Parameter Estimate SE p-value B-H critical Hedges g ES
Arithmetic
Fixed Effects
  Intercept 8.675 1.049 <.0001
  Adjusted Mean Difference
   NK v. control 7.419 1.528 <.0001 .0125 0.59
   WP v. control 7.985 1.475 <.0001 .0083 0.65
   WP[L] v. control 9.154 1.272 <.0001 .0042 0.79
   WP v. NK 0.566 1.882 .764 .0167 0.03
   WP[L] v. NK 1.735 1.496 .246 .0208 0.14
   WP v. WP[L] 1.169 1.345 .385 .0250 0.10
  Pretest effect 1.332 0.110 .001
Variance Components
  Student-level residual 99.374 6.385
  Classroom-level intercept 3.367 7.133
  School-level intercept 6.426 6.071
Word Problems
Fixed Effects
  Intercept 2.383 0.360 <.0001
  Adjusted Mean Difference
   NK v. control 0.260 0.401 .517 .0250 0.09
   WP v. control 4.045 0.526 <.0001 .0125 1.08
   WP[L] v. control 6.416 0.550 <.0001 .0042 1.75
   WP v. NK 3.785 0.576 <.0001 .0167 0.94
   WP[L] v. NK 6.156 0.563 <.0001 .0083 1.55
   WP v. WP[L] 2.371 0.684 .0001 .0208 0.47
  Pretest effect 0.632 0.171 <.0001
Variance Components
  Student-level residual 14.746 1.280
  Classroom-level intercept 0.715 1.395
Word-Problem Language
Fixed Effects
  Intercept 13.511 0.355 <.0001
  Mean Difference
   NK v. control −0.386 0.413 .349 .0250 0.17
   WP v. control 0.573 0.504 .255 .0208 0.16
   WP[L] v. control 2.164 0.559 <.0001 .0125 0.56
   WP v. NK 0.960 0.386 .013 .0167 0.24
   WP[L] v. NK 2.550 0.331 <.0001 .0042 0.63
   WP v. WP[L] 1.590 0.444 <.0001 .0083 0.41
Variance Components
  Student-level residual 13.064 1.084
  Classroom-level intercept 1.180 1.034
  School-level intercept 0.283 0.315

Note. p-values for z-tests of variance components are conservative and thus often not reported (divide est/SE and compare to +1–1.96 to discern significance according to z-test). An adjustment Fitzmaurice et al. (2011, p. 209) suggest is to employ α = .10 rather than α = .05; it performs similarly to other correction methods (Ke & Want, 2015). B-H is the Benjamini-Hochberg procedure critical value. ES is effect size.