Table 1.
All possible binary SEP permutations over three time points, expected changes in adverse health outcome BMI under different hypotheses, and corresponding linear predictor
|
SEP |
Expected value of BMI compared with the reference category of being in always manual (0,0,0), under each alternative causal model |
Regression coefficients used to calculate the predicted value of BMI under a saturated regression model shown in: |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Critical period |
Social mobility |
|||||||||
| SEP1 | SEP2 | SEP3 | Cumulative exposure* | at t1 | at t2 | at t3 | Adult | Any mobility | Equation (1) | Equation (2) |
| 0 | 0 | 0 | Reference | Reference | Reference | Reference | Reference | Reference | α | α |
| 1 | 0 | 0 | − | − | 0 | 0 | 0 | + | α+δ12D12 | α + β1S1 |
| 0 | 1 | 0 | − | 0 | − | 0 | + | 0 | α + γ12U12+ δ23D23+ψ2U12D23 | α + β2S2 |
| 0 | 0 | 1 | − | 0 | 0 | − | − | − | α + γ23U23 | α + β3S3 |
| 1 | 1 | 0 | − | − | − | 0 | + | + | α + δ23D23 | α +β1S1+β2S2+θ12S1S2 |
| 1 | 0 | 1 | - | − | 0 | − | − | 0 | α + δ12D12+γ23U23+ ψ1D12U23 | α + β1S1 + β3S3+θ13S1S3 |
| 0 | 1 | 1 | - | 0 | − | − | 0 | − | α + γ12U12 | α + β2S2 + β3S3 + θ23S2S3 |
| 1 | 1 | 1 | − | − | − | − | 0 | 0 | α + η S1 S2 S3 | α + β1S1+ β2S2+ θ3S3+ θ12S1S2+ θ23S2S3+ θ13S1S3+ θ123S1S2S3 |
Sj = 1 if SEP is non-manual at time t, Sj = 0 if SEP is manual at time t; α: expected value of Y when all Sj are 0, Dj,j+1 is a binary indicator for a downward change in social class (i.e. from Sj = 1 to Sj + 1 =0) and Uj,j + 1 is a binary indicator for an upward change (i.e. from Sj = 0 to Sj + 1 = 1). Negative sign is associated with an inverse relationship with BMI.
*The number of dashes refers to the magnitude of the inverse association between BMI and SEP.