imputing new data

[topepo]

It would be great to be able to build imputation models on data set X₁ and apply them to other data sets (X₁, X₂, ...). From what I can tell, you can only impute the same data that you start with.

[markvanderloo]

edit it is probably better to have an option to pass a model object as you suggest. I'll do that in stead :).

good point. I'm considering to add a train argument to the impute_* functions so a different training set can be used. I'll very probably do it too, even though it is considered bad practice in model-based imputation: (1) you introduce the assumption that model coefficients in the trainingset apply to the set to be imputed and (2) this means you can't do a lot of the fancy variance analyses you can do when training on the same dataset. Having said that, I think that the practice is not that bad if there's are good reasons to assume (1).

[topepo]

I'll very probably do it too

This needs to be my github signature for issues =]

My interest is more for estimation problems where I'm less concerned with the impact on the distributional assumptions. The first assumption is clearly still important though in either case. Thanks

[markvanderloo]

Max, if you install the drat version (0.2.2) there is now an impute function that works as follows:

model <- lm(foo ~ bar + baz, data=snafu)
impute(dat, imputed_variable ~ model)

code still resides in the modelimpute branch.

[markvanderloo]

now in main branch.

[topepo]

Thanks!