Skip to main content

Table 2 Response models tested

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

Response model

Set of variables

Description

1

X, Z1, Z2, Z3, Z4

All variables associated with the responsea

2

X, Z1, Z2, Z3, Z4, Z5, Z6, Z7

The exposure variable X and all covariates

3

X, Z1, Z3

The exposure variable X and variables associated with both response and outcome (strategy proposed by Hernan et al. [3])

4

X, Z1, Z2, Z3, Z5, Z7

All variables associated with the response*, except Z4 only associated with the response; Adding Z5 a confounding variable (Z5) and a prognostic variable (Z7), neither associated with the response (strategy proposed by Seaman and White [4])

5

X, Z1

The exposure variable X and the confounding variable associated with the response (Z1)

6

X, Z5

The exposure variable X and the confounding variable not associated with the response (Z5)

7

X, Z1, Z5

The exposure variable X and both confounding variables, that associated with the response (Z1) the other not (Z5)

8

X, Z1, Z5, Z7

The exposure variable X, both confounding variables (Z1, Z5) and a prognostic variable not associated with the response (Z7)

9

X, Z5, Z7

The exposure variable X and a confounding variable and prognostic variable, neither associated with the response (Z5, Z7)

10

Z1, Z2, Z3, Z4

Previous response models without the exposure variable X

11

Z1, Z2, Z3, Z4, Z5, Z6, Z7

12

Z1, Z3

13

Z1, Z2, Z3, Z5, Z7

14

Z1

15

Z5

16

Z1, Z5

17

Z1, Z5, Z7

18

Z5, Z7

  1. aX was not associated with the response in scenarios MAR 1, MNAR 1, or MNAR 4