Predicting from `Lrnr_sl` about sl3 HOT 3 CLOSED

Assuming your tmle3 fit object is named tmle3_fit, this is one way to extract the estimated propensity score from that object:
propensity_score <- tmle3_fit$likelihood$initial_likelihood$factor_list$A$learner$predict()

from sl3.

Hi @emmavestesson, thanks for bringing this up. The argument type is not an accepted argument for predict in sl3. This is why your call, sl_fit$predict(task=task, type=“response”), is producing an error. The handbook example uses sl_fit$predict(task=task). This is the correct formatting. What do the predictions look like when you run sl_fit$predict(task=task) (i.e., when you do not run the code that errors)? Could you please include, say, the first 10 predictions that are output by that call?

You can also obtain the estimated propensity scores from the tmle3 fit object. I will show you how to do that too when I to my laptop. It would be helpful in the meantime if you could please share those first few predictions so I can make sure the format of the binary outcome predictions is a predicted probability.

from sl3.

Hi @rachaelvp,

Thanks for replying so quickly, I really appreciate it. I can't share the actual values but they are all between 0-1 (not including 0 and 1). It would be great if you could share how to get the estimated propensity score from the tmle3 fit object when you have the opportunity. Thanks again for your help!

from sl3.