Odd issue with PROC MI roc mi data=f2a seed=20220807 nimpute=2 out=f3; by imp1; class trt01p oelat racegr1 region1 agegr2 serdsps1 irclgr1 trtd; var trt01p oelat racegr1 region1 agegr2 serdsps1 irclgr1 M0 M6 M12 M18 M24 M30 M36; monotone reg(M6 = M0 trt01p oelat racegr1 region1 agegr2 serdsps1 irclgr1); mnar adjust( m6 / shift=.01 adjustobs=(trtd='1') ); monotone reg(M12 = M0 M6 trt01p oelat racegr1 region1 agegr2 serdsps1 irclgr1); mnar adjust( m12 / shift=.01 adjustobs=(trtd='1') ); monotone reg(M18 = M0 M6 M12 trt01p oelat racegr1 region1 agegr2 serdsps1 irclgr1); mnar adjust( m18 / shift=.01 adjustobs=(trtd='1') ); monotone reg(M24 = M0 M6 M12 M18 trt01p oelat racegr1 region1 agegr2 serdsps1 irclgr1); mnar adjust( m24 / shift=.01 adjustobs=(trtd='1') ); monotone reg(M30 = M0 M6 M12 M18 m24 trt01p oelat racegr1 region1 agegr2 serdsps1 irclgr1); mnar adjust( m30 / shift=.01 adjustobs=(trtd='1') ); monotone reg(M36 = M0 M6 M12 M18 m24 m30 trt01p oelat racegr1 region1 agegr2 serdsps1 irclgr1); mnar adjust( m36 / shift=.01 adjustobs=(trtd='1') ); run; SAS gives me a warning WARNING: TRTD=1 is not a valid classification level in the ADJUSTOBS= option in the MNAR statement. But this only occurs when I include M36 Monotone / MNAR rows. If they are removed no warnings.   About 23% of M36 is missing and 2% of M36 has TRTD='1'   What is the issue with the M36 call?   Thanks in advance Wayne ... View more