Table 5.
Moderators of the Effect Sizes for Incident HIV and STI (k = 20).
| All comparisons |
|||
|---|---|---|---|
| Study dimension and levela | Adjustedb d+(95% CI) | β | |
| Percentage of women in sample | 0.30*** | ||
| 0% | 0.10 (0.06, 0.14) | ||
| 100% | 0.50 (0.46, 0.54) | ||
| Data collection year | 0.60*** | ||
| 1991 | −0.90 (−1.01, −0.80) | ||
| 2008 | 1.58 (1.48, 1.68) | ||
| Provided testing and treatment for STIs | 0.41*** | ||
| Absent | 0.38 (0.33, 0.43) | ||
| Present | 0.42 (0.39, 0.45) | ||
Note. Abbreviations: CI, confidence interval; d+, weighted mean effect size; STIs, sexually transmitted infections; significance of the standardized regression coefficient is denote as
p<.05;
p<.01
p<.001.
The model was a weighted least-squares multiple regression, with study dimensions simultaneously entered as independent variables and the inverse variance as the weights, following fixed-effects assumptions. Positive effect sizes imply less incident HIV/STI efficacy for the intervention group relative to the comparison group adjusted for the other variables in the model. The model explains 33% of the variance, I2(3,16)=98.68 (95% CI=98.39, 98.92).
High and low values for moderator category reflect maximum and minimum values in sample.
Holding continuous factors constant at their mean, or, in the case of provided STI testing and treatment holding this factor constant through use of contrast coding.