Table 5.
Multivariate multilevel ordinal (MCA SEP tertiles as outcome) regression analysis using a household level yearly panel data structure comparing different modelling approaches: classical, Bayesian unstructured household level random effects, and parametric distance spatial random effect.
| Model 1: Non-spatial model1 ( stata – classical fixed effects model) | Model 2: Non-spatial random effect model ( WinBUGS) | Model 3: Spatial random effect model2 ( WinBUGS) | ||||
|---|---|---|---|---|---|---|
| Factor | OR | 95% CI | OR | 95% CI | OR | 95% CI |
| Year | 1.13 | 1.12, 1.15 | 1.22 | 1.2, 1.25 | 1.23 | 1.20, 1.25 |
| Relative SEP loss (or gain) of Mozambican-headed households versus South African-headed by panel year relative to 2001 baseline | ||||||
| 2001 | 1 | – | – | |||
| 2003 | 0.78 | 0.72, 0.83 | – | – | – | – |
| 2005 | 1.01 | 0.93, 1.09 | – | – | – | – |
| 2007 | 1.03 | 0.94, 1.12 | – | – | – | – |
| Household size of 3 or more | 2.84 | 2.6, 3.09 | 3.96 | 3.52, 4.43 | 4.08 | 3.64, 4.55 |
| Household existed for at least 2 or more years before first asset observation | 1.59 | 1.35, 1.86 | 2.02 | 1.56, 2.58 | 1.72 | 1.31, 2.20 |
| Household head age of 40 or more years | 1.29 | 1.21, 1.37 | 1.5 | 1.38, 1.64 | 1.47 | 1.35, 1.60 |
| Male household head | 1.57 | 1.48, 1.67 | 2.12 | 1.91, 2.34 | 2.13 | 1.93, 2.35 |
| Mozambican household head | 0.36 | 0.33, 0.38 | 0.17 | 0.15, 0.2 | 0.20 | 0.17, 0.23 |
| Household head alive | 1 | 1 | ||||
| Died (non-HIV/AIDS) | 0.81 | 0.71, 0.92 | 0.67 | 0.54, 0.82 | 0.65 | 0.53, 0.79 |
| Died (HIV/AIDS) | 0.50 | 0.33, 0.76 | 0.32 | 0.13, 0.63 | 0.31 | 0.13, 0.63 |
| Average 4+ migrant months per household individual per year | 1.01 | 0.96, 1.07 | 1.07 | 1.00, 1.152 | 1.08 | 1.00, 1.15 |
| Average of secondary or higher education level years per household individual | 2.57 | 2.39, 2.76 | 5.31 | 4.72, 5.99 | 4.54 | 4.05, 5.09 |
| None involved in government, professional or private sector skilled work | 1 | 1 | ||||
| At least one individual involved in only one of the above | 1.83 | 1.72, 1.95 | 3.02 | 2.71, 3.36 | 2.88 | 2.59, 3.18 |
| Individuals involved in two or more of the above | 5.13 | 4.55, 5.77 | 18.70 | 15.26, 23 | 16.20 | 13.36, 19.00 |
| Constant | – | – | −3.1 | −3.37, −2.82 | −3.30 | −4.06, −2.00 |
 (household) |
– | – | 5.12 | 4.87, 5.39 | 4.71 | 4.48, 4.96 |
 (spatial) |
– | – | – | – | 1.22 | 0.62, 2.34 |
| Range (m) | – | – | – | – | 2,378 | 17,794, 269 |
| AIC( stata)/DIC( WinBUGS) | 79,282.3 | – | 61,918.9 | – | 61,670.7 | – |
MCA, multiple correspondence analysis; SEP, socio-economic position; AIC, Akaike’s information criterion; DIC, deviance information criterion.
1
Brant test of parallel regression assumption: χ2: 19.6, p > χ2: 0.108, that is, sufficient evidence to suggest that the parallel regression assumption has not been violated.
2
Model also includes an unstructured household level random effect (pooled).
3
Significant at the 10% level.