tlverse / tmle3mediate Goto Github PK

• https://github.com/nhejazi/medshift/blob/master/R/Param_stochastic_mediation.R
• https://github.com/nhejazi/medshift/blob/master/R/LF_exptilt_ipsi.R
• https://github.com/nhejazi/medshift/blob/master/R/tmle3_Spec_medshift.R

Hi, I've been trying to reproduce the mediation example on the WASH Benefits data as thoroughly explained on the tlverse website. Nonetheless, even on this example dataset, I keep getting the following error when running the tmle3 function for either NIE or NDE estimations:

> washb_NIE <- tmle3(
+   tmle_spec_NIE, washb_data, node_list, learner_list
+ )
Error in generator_fun(node_name, self$nrow) : Node outcome not specified

I am running tmle3 v0.2.0, tmle3mediate v0.0.3, and sl3 v1.4.5 (which, as far as I know, are the most recent versions). Any idea what might be causing this error? It seems to be an sl3 error, but I am not calling sl3 directly here. I did not change any code from the example.

I've uploaded a simple script in tests that when run with tmle installed locally results in a taks not found error when using tmle estimator with medshift.

I've been trying to figure out this issue.

• https://onlinelibrary.wiley.com/doi/full/10.1111/biom.12022
• https://www.degruyter.com/view/j/ijb.2012.8.issue-1/1557-4679.1361/1557-4679.1361.xml
• https://escholarship.org/uc/item/4054h5kd#page=124

Hi,

I was using tmle3mediate for a project and noted that the ATE estimated by adding the NIE and NDE together is a bit different than the estimate of the ATE you get without using tmle3mediate, using just tmle3. It would be nice if the tmle3mediate package returned the ATE w/confidence intervals as part of the object returned. Thanks!