Hi Listers,
I am using -rctmiss- in Stata v17.0 to run sensitivity analysis for missing data, when my outcome is binary (success/fail).
I am trying to better understand how the pmmdelta option works. My understanding is that by specifying -pmmdelta(1, expdelta)- is equivalent to complete case analysis, while -pmmdelta(0, expdelta)- is equivalent to assuming missing=failure. I am however unsure how I would run some alternative scenarios, still assuming that worse performance is to be observed in those with missing data.
Am I correct in assuming that exp(delta) = informative missing odds ratio (IMOR) so that if I specified -pmmdelta(0, expdelta)- I'd be assuming that those with missing data have 1/2 success rate as those with observed data; while values >1 would imply greater success in those with missing data?
Many thanks for your help!
I am using -rctmiss- in Stata v17.0 to run sensitivity analysis for missing data, when my outcome is binary (success/fail).
I am trying to better understand how the pmmdelta option works. My understanding is that by specifying -pmmdelta(1, expdelta)- is equivalent to complete case analysis, while -pmmdelta(0, expdelta)- is equivalent to assuming missing=failure. I am however unsure how I would run some alternative scenarios, still assuming that worse performance is to be observed in those with missing data.
Am I correct in assuming that exp(delta) = informative missing odds ratio (IMOR) so that if I specified -pmmdelta(0, expdelta)- I'd be assuming that those with missing data have 1/2 success rate as those with observed data; while values >1 would imply greater success in those with missing data?
Many thanks for your help!
