Table 7. Robustness Checks Using Imputed Effect Sizes (Standard Errors Reported in Parentheses).
| Non-missing effect sizes |
Missing set to zero | Largest absolute value |
Maximum effect size |
Minimum effect size |
||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Domain | Female treat. effect |
Male - female difference |
Female treat. effect |
Male - female difference |
Female treat. effect |
Male - female difference |
Female treat. effect |
Male - female difference |
Female treat. effect |
Male - female difference |
| Overall | .20** | -.03** | .17** | -.04** | .18** | -.03** | .23** | -.04** | .12* | -.03** |
| (.05) | (.01) | (.05) | (.01) | (.05) | (.01) | (.05) | (.01) | (.05) | (.01) | |
| Number of effect sizes | 676 | 808 | 808 | 808 | 808 | |||||
| Cognitive/achievement | .23** | -.03* | .19** | -.04** | .20** | -.02 | .26** | -.03** | .13 | -.03 |
| (.06) | (.02) | (.05) | (.01) | (.05) | (.01) | (.05) | (.01) | (.06) | (.01) | |
| Number of effect sizes | 441 | 556 | 556 | 556 | 556 | |||||
| Other school outcomes | -.04 | .40** | .05 | .22** | -.00 | .32** | .03 | .27** | .02 | .27** |
| (.06) | (.05) | (.06) | (.05) | (.06) | (.05) | (.06) | (.05) | (.07) | (.05) | |
| Number of effect sizes | 98 | 104 | 104 | 104 | 104 | |||||
| Child behavior/mental health | .07 | -.08* | .06 | -.07** | .07 | -.07** | .09 | -.07** | .03 | -.07** |
| (.04) | (.03) | (.04) | (.02) | (.04) | (.03) | (.04) | (.03) | (.05) | (.03) | |
| Number of effect sizes | 36 | 44 | 44 | 44 | 44 | |||||
| Adult outcomes | .18 | -.06 | .18 | -.05 | .20 | -.08 | .22 | -.07 | .17 | -.07 |
| (.10) | (.06) | (.10) | (.06) | (.10) | (.06) | (.09) | (.06) | (.12) | (.06) | |
| Number of effect sizes | 101 | 104 | 104 | 104 | 104 | |||||
Note.
represents p < .05
represents p < .01.
“Male” is a dummy variable equal to 1 for all-boys contrasts and 0 for all-girls contrasts. The female treatment effect refers to the effect size for girls. In this table, effect sizes with missing data are excluded from the analyses. Positive coefficients represent desirable outcomes, such as lower rates of special education referral or grade retention. Each cell of the table represents a separate regression using effect sizes from the given domain. Only non-significant effect sizes are imputed; significant effect sizes are set to p = .05. The four columns of imputed data mean the following: (1) Missing set to zero: All missing effect sizes were set to zero; (2) Largest absolute value: If the treatment group is favored, p = .11, if the comparison group is favored, p = .11; (3) Maximum effect size: If the treatment group if favored, p = .11, if the comparison group is favored, p = .99; (4) Minimum effect size: If the treatment group if favored, p = .99, if the comparison group is favored, p = .11.