mi and stsplit r(459) error

[Jamie Marra]

Hello:

I am working with the stcox suite. I have a large cohort and 4 covariates with missing values that were imputed using mi commands. After imputing, I split using mi stsplit at 6 failure points. It is a very large section of code so I didn't include it in this post and it runs fine most of the time. However, sometimes when I run the code I get the following output:

results could not be duplicated
    Performing stsplit on m=0 resulted in 56208 obs.  Performing the same operation on m=1 resulted in
    42084obs.  Results should have been the same.

Any idea what the problem is?

[Yulia Marchenko (StataCorp)]

The only case I can think of in which this error would arise is when your survival-time variable varies among imputations. This may happen, for example, when this variable was imputed. Stata's mi command (or mi stset, specifically), however, does not allow the survival-time variable to be one of the imputed variables. For example, if I run the code below, mi stset would issue an error message.

. sysuse auto, clear
(1978 Automobile Data)

. mi set wide

. mi register imputed mpg

. mi stset mpg
variable mpg registered as imputed
    Registered and passive variables may not be used as the basis for mi stset.
r(459);

Without actually seeing Jamie's code, however, it is hard to identify the exact origin of the error message. Jamie should contact our technical support at [email protected] to diagnose the problem.