Table 1.
Empirical coverage of the Bayesian prediction intervals for HIV prevalence
| 50% Prediction Interval | 80% Prediction Interval | 95% Prediction Interval | |||||||
|---|---|---|---|---|---|---|---|---|---|
| HIV Incidence Trend | <25% | [25%, 75%] | 75%< | <10% | [10%,90%] | 90%< | <2.5% | [2.5%, 97.5%] | 97.5%< |
| fixed gamma curve % of Observations |
10.7% | 42.9% (2.6) |
46.4% | 3.6% | 75.0% (4.9) |
21.4% | 3.6% | 92.9% (7.5) |
3.6% |
| estimated gamma curve % of Observations |
10.7% | 10.7% (1.7) |
78.6% | 3.6% | 25.0% (3.0) |
71.4% | 3.6% | 50.0% (4.2) |
46.4% |
| estimated sex-specific gamma curves % of Observations |
10.7% | 10.7% (1.8) |
78.6% | 3.6% | 28.6% (3.3) |
67.9% | 3.6% | 53.6% (4.8) |
42.9% |
| non-parametric trend % of Observations |
10.7% | 10.7% (2.0) |
78.6% | 3.6% | 39.3% (3.7) |
57.1% | 3.6% | 57.1% (5.4) |
39.3% |
| sex-specific non-parametric trends % of Observations |
10.7% | 25.0% (3.1) |
64.3% | 3.6% | 67.9% (5.7) |
28.6% | 3.6% | 89.3% (8.4) |
7.1% |
Notes: Results for Tanzania for various specifications of the HIV incidence trend. Numbers in parentheses are mean interval widths measured in percentage points. There are 28 observations.