Table 2.
Model parameter estimates derived from the Poisson regression mixed models used to evaluate the spatio-temporal variation in Manhiça child mortality rates (log standardised mortality rates; log of observed cases/expected cases) during the period 1997-2005.
| Parameter symbol | Description of parameter | Estimate | ||||
|---|---|---|---|---|---|---|
| (Standard Error) | ||||||
| Model 0* | Model 1 | Model2 | Model 3 | |||
| α0 | Baseline log-rate across all areas | 0.101 (0.051) | 0.129 (0.063) | 0.114 (0.062) | 0.114 (0.062) | |
| α1 | Log of the overall curvature trend during the period 1997-2005 | 0.009 (0.010) | 0.008 (0.010) | 0.008 (0.010) | 0.008 (0.010) | |
| α2 | -0.028 (0.004) | -0.028 (0.004) | -0.028 (0.004) | -0.028 (0.004) | ||
| β | Log-relative risks of % households with a single construction | <54% | 0.276 (0.068) | 0.231 (0.085) | 0.224 (0.084) | 0.220 (0.084) |
| 54-60% | 0.090 (0.069) | 0.052 (0.082) | 0.050 (0.082) | 0.051 (0.082) | ||
| 60-67% | -0.031 (0.067) | -0.060 (0.076) | -0.059 (0.076) | -0.056 (0.076) | ||
| >67% | baseline | baseline | baseline | baseline | ||
| Variance component reflecting the non structured variability of neighbourhood-specific intercepts | -- | 0.051 (0.017) | -- | -- | ||
| Variance component reflecting the spatially-structured variability of neighbourhood-specific intercepts | -- | -- | 0.094 (0.030) | 0.092 (0.032) | ||
| Variance components reflecting the non-structured variability of neighbourhood-specific curvature time trend | -- | -- | -- | 0.0006 (0.0013) | ||
| -- | -- | -- | 0.00008 (0.0002) | |||
| ϕ | Dispersion parameter | 1.218 | 1.067 | 1.057 | 1.036 | |
| quasiBIC | Quasi-likelihood Information criteria | 3394.66* | 2713.22 | 2704.94 | 2711.24 | |
*Model 0 is a Poisson regression model, which does not include random effects, the coefficient estimates are maximum likelihood estimates and the goodness of fit measure here is the Schwarz information criterion (BIC) based on the likelihood instead of the quasi-likelihood.