"overimpute()
This function temporarily treats each observed value in var as missing and imputes that value based on the imputation model of output. "
^^ Quote from R-Help.
The bounds parameter is used in amelia() for all NA values of a specific dataset-column.
But if the bounds parameter is set, it's not used in the function overimpute().
The problem in such a situation is, that it is possible to force amelia() in a specific way (for example to use only values between 40 and 50).
But if you use overimpute() nobody notice these boundarys.
Anyway: this behavior is missing in the documentation.
Bug, problem or wanted behavior?
"overimpute() This function temporarily treats each observed value in var as missing and imputes that value based on the imputation model of output. " ^^ Quote from R-Help.
The bounds parameter is used in amelia() for all NA values of a specific dataset-column. But if the bounds parameter is set, it's not used in the function overimpute().
The problem in such a situation is, that it is possible to force amelia() in a specific way (for example to use only values between 40 and 50). But if you use overimpute() nobody notice these boundarys.
Anyway: this behavior is missing in the documentation.
Demosource:
set.seed(1234) x.out_overimpute_bug<-amelia(africa,cs=2,ts=1 ,bounds=rbind(c(5,40,50)) ,lags="infl" ) test<-overimpute(x.out_overimpute_bug,var=c(5)) test$lower.overimputed[c(100:115)] test$mean.overimputed[c(100:115)] test$upper.overimputed[c(100:115)]
^^ values are not between 40 and 50 ..