Table 2.
Moderators of effect size.
| Moderator | k | N | g | 95% CI | Heterogeneity | ||
|---|---|---|---|---|---|---|---|
| Q | df | p | |||||
| Publication status | 1.587 | 1 | 0.208 | ||||
| Published | 17 | 1,456 | 0.404 | (0.309, 0.498) | 18.780 | 16 | 0.280 |
| Unpublished | 5 | 542 | 0.271 | (0.089, 0.454) | 2.486 | 4 | 0.647 |
| Sample ages | 2.201 | 3 | 0.532 | ||||
| General adult | 5 | 714 | 0.359 | (0.240, 0.478) | 1.866 | 4 | 0.760 |
| Children | 3 | 299 | 0.301 | (0.073, 0.530) | 1.332 | 2 | 0.514 |
| Adult patient | 4 | 326 | 0.521 | (0.258, 0.785) | 3.790 | 3 | 0.285 |
| College student | 9 | 535 | 0.447 | (0.272, 0.622) | 10.705 | 8 | 0.269 |
| Goal domain | 3.075 | 3 | 0.380 | ||||
| Academic | 5 | 414 | 0.255 | (0.062, 0.449) | 5.731 | 4 | 0.220 |
| Health | 12 | 1,154 | 0.379 | (0.273, 0.485) | 10.389 | 11 | 0.496 |
| Personal | 4 | 350 | 0.457 | (0.241, 0.673) | 3.658 | 3 | 0.301 |
| Relationship | 1 | 80 | 0.609 | (0.160, 1.058) | 0 | 0 | 1.000 |
| Type of intervention | 4.797 | 1 | 0.029 | ||||
| Document intervention | 6 | 702 | 0.277 | (0.154, 0.399) | 6.617 | 5 | 0.251 |
| Experimenter intervention | 16 | 1,296 | 0.465 | (0.349, 0.580) | 11.439 | 15 | 0.721 |
| Dependent measure | 1.403 | 1 | 0.236 | ||||
| Objective | 5 | 459 | 0.272 | (0.078, 0.467) | 3.878 | 4 | 0.423 |
| Self-report | 15 | 1,452 | 0.403 | (0.308, 0.499) | 17.386 | 14 | 0.236 |
| Expectation level | 1.466 | 1 | 0.226 | ||||
| High | 5 | 429 | 0.297 | (0.097, 0.496) | 9.099 | 4 | 0.059 |
| Low | 4 | 319 | 0.481 | (0.258, 0.705) | 1.221 | 3 | 0.748 |
k = number of studies, N = sample size, Q represents Q heterogeneity test. Publication status: published means published articles; unpublished means unpublished thesis and dissertation. In the sample ages, three studies were excluded because one study consisted of both children and adults, and two studies did not report the age of the sample. In dependent measures, 1 study was excluded because this study consisted of both self-report and objective measures. At the expectation level, we used expectations of success (measured by “how likely do you think it is you will…,” Likert 7-point scales) as indicators and defined 5 and above as high expectation and lower than 5 as low expectation. Thirteen studies were excluded because they did not report expectations of success. Two independent studies by Kizilcec and Cohen (2017) were excluded from moderator analysis.