Table 1

 Hypothetical multilevel regression analysis of systolic blood pressure (SBP) (mm Hg) in 25 000 people aged 35 to 64 from 39 neighbourhoods of a city

Model with individual variables and complex variance (iv0)Model with neighbourhood variables (v)Model with neighbourhood variables (v°)
Fixed effects and random effects are expressions often used in MLRA. In very simple terms, fixed effects are used to model means, whereas random effects are used to model variances. *Regression coefficients (β) and 95% confidence intervals (95% CI). BMI, body mass index centred on the mean; PCV, proportional change in variance; VPC, variance partition coefficient. †The reference group is people living in advantaged neighbourhoods and with BMI = 25 kg/m2. ‡In people with BMI = 25 kg/m2.
Fixed effects(Traditional measures of association)
Mean SBP of the city (intercept)119 mm Hg116 mm Hg116 mm Hg
β* (95% CI) β * (95% CI) β * (95% CI)
Individual level variables
    Age (years)0.76 (0.70 to 0.82)0.76 (0.70 to 0.82)0.76 (0.70 to 0.82)
    Use of blood pressure lowering drugs (yes v no)0.89 (0.77 to 1.01)0.88 (0.76 to 1.00)1.02 (0.88 to 1.16)
    BMI (kg/m2)15.12 (14.43 to 15.81)15.12 (14.43 to 15.81)15.12 (14.43 to 15.81)
    Individual low income (low v high)2.27 (1.72 to 2.82)2.27 (1.72 to 2.82)2.26 (1.71 to 2.81)
Area level variables
    Neighbourhood low income (low v high)5.23 (3.19 to 7.27)4.47 (2.49 to 6.45)
Cross level interactions
    Neighbourhood low income × BMI†0.35 (0.15 to 0.55)
Random effects(Multilevel measures of health variation)
Components of variance Variance‡(SE) Variance‡(SE) Variance‡(SE)
Neighbourhood intercept variance27.09 (6.26)21.11 (4.7)21.02 (4.9)
Individual level variance298.22 (3.34)298.03 (3.34)298.02 (3.34)
BMI-SBP slope variance0.11 (0.03)0.11 (0.03)0.09 (0.03)
BMI related individual SBP variance0.16 (0.08)0.16 (0.08)0.16 (0.08)
Neighbourhood intercept-slope covariance0.99 (0.35)0.62 (0.29)0.58 (0.27)
Individual level covariance5.24 (0.31)5.24 (0.31)5.25 (0.31)
Proportional change in variance (PCV)PCVPCVPCV
    PCV between neighbourhoodsRef.22%22%
    PCV between individualsRef.0%0%
Variance partition coefficient (VPC)VPC  =  0.08VPC = 0.07VPC = 0.07
2×log-likelihood value211,024211,006210,996