Graphical models for inference under outcome-dependent sampling

TY - JOUR

T1 - Graphical models for inference under outcome-dependent sampling

AU - Didelez, V

AU - Kreiner, S

AU - Keiding, N

PY - 2010/8

Y1 - 2010/8

N2 - We consider situations where data have been collected such thatthe sampling depends on the outcome of interest and possibly further covariates,as for instance in case-control studies. Graphical models representassumptions about the conditional independencies among the variables. Byincluding a node for the sampling indicator, assumptions about samplingprocesses can be made explicit. We demonstrate how to read off such graphswhether consistent estimation of the association between exposure and outcomeis possible. Moreover, we give sufficient graphical conditions for testingand estimating the causal effect of exposure on outcome. The practicaluse is illustrated with a number of examples.

AB - We consider situations where data have been collected such thatthe sampling depends on the outcome of interest and possibly further covariates,as for instance in case-control studies. Graphical models representassumptions about the conditional independencies among the variables. Byincluding a node for the sampling indicator, assumptions about samplingprocesses can be made explicit. We demonstrate how to read off such graphswhether consistent estimation of the association between exposure and outcomeis possible. Moreover, we give sufficient graphical conditions for testingand estimating the causal effect of exposure on outcome. The practicaluse is illustrated with a number of examples.

U2 - 10.1214/10-STS340

DO - 10.1214/10-STS340

M3 - Journal article

SN - 0883-4237

VL - 25

SP - 368

EP - 387

JO - Statistical Science

JF - Statistical Science

IS - 3

ER -