supporting doubly robust inference

[nhejazi]

we should also support David's very general drtmle package, which is on CRAN and on GitHub. This can come in a subsequent release on Bioconductor.

This has mostly been implemented on the branch drtmle. There remain a few (possibly) outstanding issues that need to be dealt with

• drtmle (sensibly) does not support the setting where W = 1 uniformly. There are suggestions on how to handle this setting here
• A recent update of drtmle on CRAN (v1.0.2) returns the observed data after application of the efficient influence function transform. This makes it possible to port the variance shrinkage approach (based on limma and implemented in biotmle) to this setting.

[nhejazi]

This has mostly been implemented on the branch drtmle. There remain a few (possibly) outstanding issues that need to be dealt with

• drtmle (sensibly) does not support the setting where W = 1 uniformly. There are suggestions on how to handle this setting here
• A recent update of drtmle on CRAN (v1.0.2) returns the observed data after application of the efficient influence function transform. This makes it possible to port the variance shrinkage approach (based on limma and implemented in biotmle) to this setting.