Table 2.
Task Moderator Analyses
| Moderator and levels | j | k | g | β | 95% CI | p | R (2) 2 | R (3) 2 | Q | τ(2)2 | τ(3)2 |
|---|---|---|---|---|---|---|---|---|---|---|---|
|
| |||||||||||
| Test type (recognition vs. recall) | 233 | 988 | — | −0.17 | [−0.27, −0.06] | .002 | .058 | .007 | 3504.2* | 0.08* | 0.11* |
| Recognition | 110 | 342 | −0.24 | — | [−0.32, −0.16] | — | — | — | — | — | — |
| Recall | 188 | 646 | −0.41 | — | [−0.47, −0.34] | — | — | — | — | — | — |
| Recall type (cued vs. free) | 186 | 643 | — | −0.04 | [−0.15, 0.06] | .427 | .009 | .000 | 2429.9* | 0.09* | 0.13* |
| Cued | 40 | 80 | −0.37 | — | [−0.48, −0.25] | — | — | — | — | — | — |
| Free | 173 | 563 | −0.41 | — | [−0.48, −0.35] | — | — | — | — | — | — |
| Test delay | 217 | 921 | — | — | — | — | .000 | .004 | 3238.9* | 0.09* | 0.10* |
| Immediate vs. short delay | 194 | 660 | — | 0.02 | [−0.06, 0.09] | .679 | — | — | — | — | — |
| Immediate vs. long delay | 170 | 644 | — | 0.01 | [−0.04, 0.07] | .665 | — | — | — | — | — |
| Short delay vs. long delay | 168 | 538 | — | −0.004 | [−0.08, 0.07] | .922 | — | — | — | — | — |
| Immediate | 140 | 380 | −0.38 | — | [−0.43, −0.31] | — | — | — | — | — | — |
| Short delay | 92 | 277 | −0.35 | — | [−0.43, −0.28] | — | — | — | — | — | — |
| Long delay | 108 | 261 | −0.36 | — | [−0.42, −0.30] | — | — | — | — | — | — |
| Stimili material | 229 | 978 | — | — | — | — | .000 | .044 | 3482.7* | 0.09* | 0.10* |
| Words vs. high verbal | 223 | 898 | — | 0.06 | [−0.05, 0.17] | .266 | — | — | — | — | — |
| Words vs. low verbal | 206 | 827 | — | −0.10 | [−0.21, 0.01] | .074 | — | — | — | — | — |
| High vs. low verbal | 78 | 231 | — | −0.17 | [−0.30, −0.03] | .024 | — | — | — | — | — |
| Verbal words | 200 | 747 | −0.36 | — | [−0.41, −0.30] | — | — | — | — | — | — |
| Verbal pictures | 44 | 151 | −0.29 | — | [−0.40, −0.19] | — | — | — | — | — | — |
| Nonverbal | 44 | 80 | −0.46 | — | [−0.57, −0.35] | — | — | — | — | — | — |
| Valence | 232 | 984 | — | — | — | — | .071 | .186 | 3470.7* | 0.08* | 0.09* |
| Neutral vs. positive | 231 | 815 | — | 0.03 | [−0.06, 0.12] | .516 | — | — | — | — | — |
| Neutral vs. negative | 231 | 840 | — | 0.33 | [0.23, 0.43] | <.001 | — | — | — | — | — |
| Positive vs. negative | 87 | 313 | — | 0.30 | [0.17, 0.44] | <.001 | — | — | — | — | — |
| Neutral | 200 | 671 | −0.42 | — | [−0.47, −0.36] | — | — | — | — | — | — |
| Positive | 75 | 144 | −0.39 | — | [−0.48, −0.30] | — | — | — | — | — | — |
| Negative | 86 | 169 | −0.08 | — | [−0.18, 0.01] | — | — | — | — | — | — |
| Memory outcome | 236 | 995 | — | — | — | — | .037 | .029 | 3524.9* | 0.09* | 0.10* |
| Accuracy vs. discrimination | 234 | 904 | — | 0.08 | [−0.19, 0.36] | .553 | — | — | — | — | — |
| Accuracy vs. com errors | 222 | 897 | — | 0.22 | [0.08, 0.36] | .004 | — | — | — | — | — |
| Discrimination vs. com errors | 69 | 189 | — | 0.13 | [−0.16, 0.43] | .375 | — | — | — | — | — |
| Accuracy | 220 | 806 | −0.38 | — | [−0.45, −0.32] | — | — | — | — | — | — |
| Discrimination | 32 | 98 | −0.30 | — | [−0.55, −0.05] | — | — | — | — | — | — |
| Errors | 38 | 91 | −0.16 | — | [−0.30, −0.03] | — | — | — | — | — | — |
| Memory process (Fam. vs. Rec.) | 11 | 50 | — | −0.25 | [−0.66, 0.15] | .274 | .146 | .000 | 146.6* | 0.07* | 0.10 |
| Familiarity | 10 | 23 | −0.18 | — | [−0.45, 0.09] | — | — | — | — | — | — |
| Recollection | 11 | 27 | −0.44 | — | [−0.79, −0.09] | — | — | — | — | — | — |
| Publication status (unpub vs. pub) | 236 | 995 | — | −0.19 | [−0.33, −0.06] | .008 | .000 | .050 | 3524.9* | 0.09* | 0.10* |
| Unpublished | 37 | 210 | −0.20 | — | [−0.32, −0.08] | — | — | — | — | — | — |
| Published | 199 | 785 | −0.39 | — | [−0.45, −0.34] | — | — | — | — | — | — |
| Valence in self-reference tasks (positive vs. negative) | 34 | 99 | — | 0.71 | [0.38, 1.04] | <.001 | .293 | 0 | 574.4* | 0.03* | a |
| Positive | 31 | 47 | −0.45 | — | [−0.66, −0.24] | — | — | — | — | — | — |
| Negative | 34 | 52 | 0.26 | — | [0.10, 0.43] | — | — | — | — | — | — |
| Valence in external focus tasks (positive vs. negative) | 57 | 204 | — | 0.12 | [−0.003, 0.24] | .064 | .027 | .002 | 718.4* | 0.11* | 0.06* |
| Positive | 46 | 92 | −0.23 | — | [−0.35, −0.11] | — | — | — | — | — | — |
| Negative | 56 | 112 | −0.11 | — | [−0.21, −0.01] | — | — | — | — | — | — |
Note. The moderator name is in boldface, the specific comparison between levels is in boldface italics. The first variable in the comparisons serves as the baseline (e.g., recognition is the baseline). j = number of unique depression-control group comparisons. k = number of effect sizes. g = Hedges’ g. β coefficients are from metaregression analyses where categorical moderators with two or three levels were dummy coded and entered into the models as predictors. 95% CI corresponds to the β for moderators or g values for individual levels of moderators. p corresponds to the β for moderators. R(2)2 = proportion of estimated heterogeneity explained by the predictors at Level 2. R(3)2 = proportion of estimated heterogeneity explained by the predictors at Level 3. Q = Q statistic on the homogeneity of effect sizes. τ(2)2 = heterogeneity of effects due to differences between measures. τ(3)2 = heterogeneity of effect sizes across studies after controlling for the different types of measures at Level 2. High verbal = high verbalizable stimuli; Low verbal = low verbalizable stimuli; Com errors = commission errors; Fam. = familiarity; Rec. = recollection; Unpub = unpublished; Pub = published. For the levels of the moderators, j does not necessarily add up to the total j for the moderator (nor do the percentages add up to 100) because depression and control group performance may be compared on both recall and recognition, for example, within a single study.
The three-level model indicated there was not enough variance at Level 3, creating a convergence issue. However, rerunning this analysis with a two-level model yielded the same result as the three-level model.
p < .05 for Q and τ2 statistics.