Table 4.

Effect Sizes as a Function of Moderator Variables

Moderator	Outcome
	Letter name knowledge		Letter sound knowledge		Letter writing		Letter name fluency
	k	Effect size	k	Effect size	k	Effect size	k	Effect size
Use of random assignment
Yes	25	0.40	26	0.65	1	0.915	12	0.18 *
No	13	0.31	16	0.56	6	0.497	6	−0.37
Presence of selection bias(es)
Yes	24	0.33	22	0.51	5	0.497	12	−0.09
No	14	0.43	20	0.74	2	0.746	6	0.21
Alphabet instruction isolated to the treatment group
Yes	35	0.38	37	0.61	7	0.545	16	0.04
No	3	0.20	5	0.63	0		2	−0.24
Potential ceiling effects
Yes	11	0.47	9	0.53	3	0.705
No	19	0.35	25	0.68	4	0.427
Publication bias
Peer-reviewed journal	24	0.46	35	0.68	3	0.336	8	−0.08
Other source	14	0.22	6	0.45	4	0.662	10	0.07
Participant age
Early childhood  (preschool/kindergarten)	33	0.37	35	0.65	5	0.601	14	0.09
Elementary	5	0.34	6	0.43	2	0.381	3	0.32
Participants at risk for reading difficulties
Yes	26	0.40	18	0.42	2	0.569	7	−0.14
No	2	0.58	4	0.57	0		2	0.28
Instructional setting
School	34	0.41 #	35	0.63 #	6	0.549	15	0.11
Home	2	−0.80 #	2	−0.61 #	0		2	0.24
Lab or clinic	1	0.00	1	0.78	0		0
Implementer
Teacher	18	0.54	16	0.63	5	0.594	9	0.16
Research staff	16	0.31	23	0.71	1	0.5	5	0.01
Group size
Individual	4	−0.26 #	9	0.39	2	0.355	4	0.30
Small group	21	0.52 #	20	0.73	2	0.569	7	0.07
Whole class	12	0.24	11	0.48	3	0.607	5	0.06
Instructional timea
Minimum	30	−0.26 *	32	0.57	3	0.508	15	0.11
Mean		0.31		0.50		0.546		0.15
Max		0.98		0.42		0.565		0.19

Note. Contrasts represent results of pairwise comparisons within factors.

^aTotal minutes of instructional time was highly positively skewed, and a logarithmic transformation of this variable was used in all analyses (Tabachnick & Fidell, 1996). Because this moderator was a continuous variable, effects are listed for instructional time at the minimum, mean, and maximum.

^*Contrast between effect sizes is significant at p < .05.

^#Contrast between effect sizes trended towards significance at p < .10 and significantly improved model fit at p < .05.