Skip to main content
. 2014 Nov 10;109(12):1986–1993. doi: 10.1111/add.12721

Table 5.

Results from the sensitivity analysis for Hedeker's data. The notation treatment/control in the Assumption column refers to the assumption made in the treatment and control groups, respectively

β2 β4 β6 β8 Assumption α (Standard error) OR (95% CI)
MAR (ignoring X; section 3.3) −0.35 (0.26) 0.71 (0.43,1.17)
0 0 0 0 MAR (using X and Z) −0.33 (0.25) 0.72 (0.44, 1.18)
−∞ +∞ −∞ +∞ LOCF −0.39 (0.22) 0.68 (0.44, 1.03)
+∞ +∞ +∞ +∞ ‘Missing = smoking’ −0.48 (0.25) 0.62 (0.38, 1.01)
0 0 −∞ +∞ MAR/LOCF −0.21 (0.23) 0.81 (0.51, 1.28)
0 0 +∞ +∞ MAR/'missing = smoking’ −0.74 (0.25) 0.48 (0.29, 0.78)
−∞ +∞ 0 0 LOCF/MAR −0.51 (0.24) 0.60 (0.38, 0.95)
−∞ +∞ +∞ +∞ LOCF/'missing = smoking’ −0.92 (0.23) 0.40 (0.25, 0.63)
+∞ +∞ 0 0 ‘Missing = smoking'/MAR −0.08 (0.25) 0.93 (0.57, 1.52)
+∞ +∞ −∞ +∞ ‘Missing = smoking'/LOCF 0.05 (0.23) 1.05 (0.67, 1.65)

MAR = missing at random; LOCF = last observation carried forward; OR = odds ratio; CI = confidence interval.