olivierlabayle
commented
3 months ago Thank you @roskamsh, I've added a few comments. Also, I think we should drop the following part:
" There are two main ways to define targeted estimators, from a predefined configuration or from a custom file.
Predefined estimators
A set of predefined estimators is readily available and can be accessed by using the configuration's name as the ESTIMATOR_FILE parameter. For instance, we provide the following:
- G-superlearning-Q-glm
- G-superlearning-Q-glmnet
- glm-with-interactions-for-Q
- glm
- glmnet-with-interactions-for-Q
- glmnet
- superlearning-with-interactions-for-Q
- superlearning
- tuned-xgboost
To see what the estimand files for these examples look like, please refer to the TargetedEstimation.jl package, located here: https://github.com/TARGENE/TargetedEstimation.jl/tree/main/estimators-configs.
All using TMLE as the meta statistical inference method. While these should cover most use cases, it may be useful to define a custom estimation strategy. This can be achieved by writing a small Julia file described below. "
since this is not working atm... Unless you are willing to give it a go :)
The purpose of this branch is to update the documentation with some current recommendations for estimators to use. These cover two scenarios - (1) ATE estimation and (2) IATE estimation. This is with the caveat that these are likely to be amended un a future release of Targene.