Table 4. Regression models for the log of the under–5 mortality rate for different model specifications for 188 countries 1970–2013.

	Coefficient SE		95% CI
Lagged distributed income (logarithmic scale)
Mixed effects regression	−0·16‡	0·007	−0·18 to −0·15
Within–between regression*	−0·15‡	0·013	−0·17 to −0·12
Generalised linear model	−0·15‡	0·007	−0·17 to −0·14
Within–between regression†	−0·15‡	0·008	−0·17 to −0·14
Maternal education
Mixed effects regression	−0·08‡	0·004	−0·09 to −0·08
Within–between regression*	−0·09‡	0·009	−0·11 to −0·07
Generalised linear model	−0·08‡	0·004	−0·09 to −0·07
Within–between regression†	−0·08‡	0·004	−0·09 to −0·07
Crude death rate from HIV
Mixed effects regression	92·42‡	4·137	84·31 to 100·52
Within–between regression*	56·42‡	4·323	47·95 to 64·90
Generalised linear model	91·26‡	4·098	83·22 to 99·29
Within–between regression†	91·26‡	4·156	83·13 to 99·38

Combined GBD super–region and year fixed effects, and country level random effects, when included, not shown here.

^*Within–between estimator with AR(1) autocorrelation specification.

^†Within–between estimator without AR(1) autocorrelation specification.

^‡Significant at 0·001 level.