Skip to main content

Table 3 Summary of imputation results for self-report abstinence assuming different levels of association between the survey missing status and self-report abstinence

From: Analysis of self-report and biochemically verified tobacco abstinence outcomes with missing data: a sensitivity analysis using two-stage imputation

  Counseling vs. no counseling Multiple vs. single contests
OR 1 Counseling Tx2 + Tx4 No counseling Tx1 + Tx3 Estimated treatment effect (odds ratio for abstinence) P-value Multiple contests
Tx3 + Tx4
Single contest
Tx1 + Tx2
Estimated treatment effect (odds ratio for abstinence) P-value
Complete case only 29.7% 24.4% 1.31 .058 28.4% 25.4% 1.16 .291
 1 29.8% 24.4% 1.31 .034 28.6% 25.4% 1.18 .212
 2 27.0% 22.7% 1.26 .086 26.2% 23.4% 1.16 .251
 3 25.8% 22.0% 1.23 .125 25.2% 22.5% 1.16 .275
 4 25.1% 21.7% 1.21 .154 24.7% 22.1% 1.15 .290
 5 24.7% 21.5% 1.20 .175 24.3% 21.8% 1.15 .301
 ⁞         
 +∞ 22.9% 20.5% 1.15 .303 22.8% 20.6% 1.14 .359
  1. OR1: odds ratio between missing and tobacco use status for self-report data, where OR1 = 1 corresponds to the situation when missing is independent of tobacco use and OR1 = positive infinity (+∞) corresponds to the situation when missing = smoking; Tx1: single contest + no counseling; Tx2: single contest + counseling; Tx3: multiple contests + no counseling; Tx4: multiple contest + counseling. P-values are based on the Chi-square test