Skip to main content

Table 1 Response mechanism scenarios (data generation)

From: Inverse probability weighting to handle attrition in cohort studies: some guidance and a call for caution

Scenario

γx

γy

\({\boldsymbol{\gamma}}_{\mathbf{1}},{\boldsymbol{\gamma}}_{\mathbf{2}},{\boldsymbol{\gamma}}_{\mathbf{3}},{\boldsymbol{\gamma}}_{\begin{array}{c}\mathbf{4}\\ {}\ \end{array}}\)

Description

MAR 1

0.0

0.0

0.1

Response depending only on covariates

MAR 2

0.2

0.0

0.1

Response depending on covariates and exposure

MAR 3

0.5

0.0

0.1

Response depending on covariates and exposure

MNAR 1

0.0

0.2

0.1

Response depending on outcome and covariates

MNAR 2

0.2

0.2

0.1

Response depending on outcome, exposure, and covariates

MNAR 3

0.5

0.2

0.1

Response depending on outcome, exposure, and covariates

MNAR 4

0.0

0.5

0.1

Response depending on outcome and covariates

MNAR 5

0.2

0.5

0.1

Response depending on outcome, exposure, and covariates

MNAR 6

0.5

0.5

0.1

Response depending on outcome, exposure, and covariates

  1. γk:regression coefficients of the generated response models (logit(pi) = γ0 + γy yi + γx xi + γ1 z1i + γ2 z2i + γ3 z3i + γ4 z4i)