Hi Listers,
I am planning to use -rctmiss- with a binary outcome. I read that when specifying a pattern-mixture model, the delta can be interpreted as an informative missing odds ratio (IMOR) = positive odds in missing data / positive odds in observed data so that delta = 1 corresponds to missing = fail (as in the smoking example, missing = smoking).
I wanted to check how to best interpret other delta values for binary outcomes; if I am expecting fewer positive odds in the missing data, should I use delta values <1, for example: delta = 0.1?
Can these delta values be interpreted as follows; if delta = 0.1, those with missing data are 90% less likely (than those with non-missing data) to succeed/quit smoking?
Thank you!
I am planning to use -rctmiss- with a binary outcome. I read that when specifying a pattern-mixture model, the delta can be interpreted as an informative missing odds ratio (IMOR) = positive odds in missing data / positive odds in observed data so that delta = 1 corresponds to missing = fail (as in the smoking example, missing = smoking).
I wanted to check how to best interpret other delta values for binary outcomes; if I am expecting fewer positive odds in the missing data, should I use delta values <1, for example: delta = 0.1?
Can these delta values be interpreted as follows; if delta = 0.1, those with missing data are 90% less likely (than those with non-missing data) to succeed/quit smoking?
Thank you!
